The Whys of Subnuclear Physics - Zichichi (Ed)

The lIftys of
Subnuclear Physics
THE SUBNUCLEAR SERIES
Series Editor: ANTONI NO ZICHICHI

European Physical Society
Geneva, Switzerland
I. 1963 STRONG, ELECTROMAGNETIC, AND WEAK INTERACTIONS
2. 1964 SYMMETRIES IN ELEMENTARY PARTICLE PHYSICS
3. 1965 RECENT DEVELOPMENTS IN PARTICLE SYMMETRIES
4. 1966 STRONG AND WEAK INTERACTIONS
5. 1967 HADRONS AND THEIR INTERACTIONS
6. 1968 THEORY AND PHENOMENOLOGY IN PARTICLE PHYSICS
7. 1969 SUBNUCLEARPHENOMENA
8. 1970 ELEMENTARY PROCESSES AT HIGH ENERGY
9. 1971 PROPERTIES OF THE FUNDAMENTAL INTERACTIONS
10. 1972 HIGHLIGHTS IN PARTICLE PHYSICS
II. 1973 LAWS OF HADRONIC MATTER
12. 1974 LEPTON AND HADRON STRUCTURE
13. 1975 NEW PHENOMENA IN SUBNUCLEAR PHYSICS
14. 1976 UNDERSTANDING THE FUNDAMENTAL CONSTITUENTS

OF MATTER
15. 1977 THE WHYS OF SUBNUCLEAR PHYSICS
Volume 1 was published by W. A. Benjamin, Inc., New York; 2-8 and 11-12 by Academic Press,
New York and London; 9-10 by Editrice Compositori, Bologna; 1315 by Plenum Press, New York
and London.
The lIflys of
Subnucleor Physics
Edited by
Antonino Zichichi
European Physical Society
Geneva, Switzerland
PLENUM PRESS NEW YORK AND LONDON

Library of Congress Cataloging in Publication Data
International School of Subnuclear Physics, Erice, Italy, 1977.
The whys of subnuclear physics.
(The subnuclear series; 15)
"Proceedings of the 1977 International School of Subnuclear Physics,
held in Erice, Trapani, Sicily, July 23-August 10, 1977."
Includes index.
1. Particles (Nuclear physics)-Congresses.
I. Zichichi, Antonino. II. Title. III. Series.
QC793.I5551977a 539.7'21 78-31740
ISBN 978-1-4684-0993-2 ISBN 978-1-4684-0991-8 (eBook)
DOI 10.1007/978-1-4684-0991-8
Proceedings of the 1977 International School of Subnuclear Physics

held in Erice, Trapani, Sicily, July 23-August 10, 1977
1979 Plenum Press, New York

Softcover reprint of the hardcover 1st edition 1979
A Division of Plenum Publishing Corporation
227 West 17th Street, New York, N.Y. 10011
All rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted,

in any form or by any means, electronic, mechanical, photocopying, microfilming,
recording, or otherwise, without written permission from the Publisher
Preface
From 23 July to 10 August 1977 a group of 125 physicists from

72 laboratories of 20 countries met in Erice to attend the 15th
Course of the International School of Subnuclear Physics.
The countries represented at the School were: Belgium, Bulgaria,

Denmark, Federal Republic of Germany, Finland, France, Hungary,
Ireland, Israel, Italy, Japan, the Netherlands, Norway, Poland,
Sweden, Switzerland, the United Kingdom, the United States of
America and Venezuela.
The School was sponsored by the Italian Ministry of Public

Education (MPI), the Italian Ministry of Scientific and Technologi-
cal Research (MRST) , the North Atlantic Treaty Organization (NATO),
the Regional Sicilian Government (ERS) and the Heizmann Institute
of Science.
The School was very exciting due to the impressive number of

frontier problems which were discussed. Being the 15th year of the
School, it was decided to review all outstanding "Whys".
At various stages of my work I have enjoyed the collaboration

of many friends whose contributions have been extremely important
for the School and are highly appreciated. I would like to thank
Dr.A. Gabriele, Ms.S. McGarry, Mr. and Mrs. S. Newman, Ms.P. Savalli
and Ms.M. Zaini for the general scientific and administrative work.
Finally, I would like to thank most warmly all those ~n Erice,

Bologna and Geneva who helped me on so many occasions and to whom I
feel very much indebted.
A. zichichi
Geneva, September 1978
v
Pn06eh~on Sidn~y Coleman h~ b~~n awand~d ~h~
P~z~ 06
"Beh~ L~c~lUt~n"
onW~ occ~ion 06~h~ 15w anniv~any 06~h~

I~~national School 06 Su.bnu.u~an Phy~i~.
Contents
The Hhys of Subnuclear Physics 1

A. Zichichi
Why Is There Charm, Strangeness, Colour

and All That? ............................ 11
H. J. Lipkin
Do Mesons Fill SU(3) Nonets? 159

R. Bizzarri
The Properties of Charmonium and Charm

Particles .................... 203
H. Schopper
Recent Results from DASP 357

B. H. Hiik
New Particles or "Why I Believe in Quarks" 395

A. Martin
New Particle Production in Hadronic Interactions 435

M. Chen
Review of Lepton Production in Hadron - Hadron

Collisions ............................... 471
A. J. S. Smith
Narrow Resonances in BB Reactions 533

S. Nilsson
Parton Distributions and Their Q2 Dependence 581

N. Cabibbo
Total Cross Sections of Neutrinos and

Antineutrinos in BEBC in
the Energy Range 20-200 GeV 607
P. Renton
vii
viii CONTENTS
Measurement of Neutral Current Cross-

Sections and Their Energy
and y - Dependence 635
H. P. Paar
Charged V+A Currents in Left-Right Symmetric

Ga uge Models 671
R. Bunny
Quark and Lepton Mixing 691

N. Cabibbo
Quark-Geometrodynamics: A New Approach to

Hadrons and Their Interactions 727
G. Preparata
The Uses of Instantons 805

S. Coleman
Can ~le Make Sense Out of "Quantum

Chromodynamics"? 943
G. 't Hooft
Should We Believe in Quantum Field Theory? 983

A. S. Wightman
An Exact Relativistic S-Matrix in 1+1 Dimensions:

The On-Shell Solution of the
}f.assive Thirring Hodel and the
Quantum Sine-Gordon Equation 1027
M. Karowski
Dynamical Symmetries in Nuclear Physics 1043

F. Iachello
Symmetries of Quarks and Leptons 1059

F. Gursey
The Best Why . 1165

A. Sanda
Status of the Subnuclear ~fuys 1171

A. Zichichi
List of Prizes Awarded, Scholarships, etc. 1215
List of Participants 1217
Index................................................... 1231
THE WHYs OF SUBNUCLEAR PHYSICS
Antonino Zichichi
CERN
Geneva, Switzerland
In the last few years an impressive series of spectacular re-

sults have been obtained in high-energy physics: notably, the dis-
coveries of the neutral weak currents, of the new particles (J,~'s),
of the new lepton with 'V l.9 GeV mass, of the "new physics" in the
(3.9-4.5) GeV mass range of SPEAR, and of the dileptons plus the
electron-strange particle production induced by high-energy v .
]J
In spite of these spectacular results, subnuclear physics is

far from reaching the asymptotic limit of a field without future.
This &s testified by the large number of problems which arise

at a rate at least comparable with that of the new results.
The purpose of this lecture is to review these problems. Let

us start with "isospin".
1. DO WE UNDERSTAND ISOSPIN?
i) Why is isospin such a good symmetry? If there are four

or more quarks, why does the quark mass spectrum begin
with one nearly degenerate doublet (u,d) while all the
remaining states are singlets with large mass separa-
tions?
2 A. ZICHICHI
ii) Why do degenerate isoscalar-isovector doublets occur fre-

quently in the hadron mass spectrum (A-L, p-w, f-A 2 )?
iii) Why are the pseudoscalar mesons so different, with no

isoscalar state degenerate with the pion?
2. DO WE UNDERSTAND STRANGENESS?
i) Why are total cross-sections for strange particles smal-

ler than for non-strange particles? Total cross-sections
of strange particles are smaller than the corresponding
non-strange ones; strange particles are heavier than the
non-strange ones. Are those two facts correlated? Also
:::: 2-3 mb 0
TIp
~ 3.1 GeV m
TI
ii) Why are the weak non-leptonic decay rates three orders of
magnitude greater than the weak leptonic decay, i.e.
A -+ S
?
A -+ all
iii) Why are S- and P-waves in non-leptonic strange particle

decays correlated?
Strange particles behave differently from non-strange particles

in ways which are still not understood. The assertion that strange
and non-strange particles differ because of the strange-quark con-
tent, merely passes the buck to the quark level and explains no-
thing. Do we understand strangeness?
3. DO WE UNDERSTAND THE NEW PARTICLES?
i) There are four states with C = +1, between J and ~'. Do

we understand their decay rates?
ii) In the hidden charm states, why is the triplet-singlet

splitting so large? Is the answer: long-range spin-
dependent forces?
THE WHYS OF SUBNUCLEAR PHYSICS 3
iii) Is the J PC = 1+- state inside the mass-range 3.550. -

3.4l5? or is it pushed down by a large triplet-singlet
shift?
iv) Charmed mesons, D, DO, D*O, D*, have been observed,

with the expected cos 2 8 behaviour and parity violation.
c
Is the open colour theory still alive?
v) At 4.0.28, analysis of the recoil mass spectrum against

the detected DO's indicates that

a(D_0 D ): a(D-0 D*0 + -*0
D D): a(D-*0 :D *0 ) ~ 1:8:11 .
This yields the following ratios of the reduced coupling

constants:
g2 (DD):g 2 (DD *):g 2 (D *D~< ) ~ 1:5:10.0.
Is the answer: Holecular charmonium?
vi) The charm yield at 4.415 GeV is 'V 1/3 that at 4.0.28 GeV.
Is this due to charm burning?
vii) \lliy does KO -7- 11+11- not go as it should, while vp -7- v +

L
anything does? o.r more generally, why are weak neutral
non-strange currents there, and weak neutral strange cur-
rents not there? Is the answer: charm, heavy leptons,
or something else?
viii) Should we believe in the spikes observed ~n pp annihila-

tions?
4. DO. WE UNDERSTAND SYMMETRY BREAKING?
i) \lliy is the Cabibbo angle 'V 2o.O? Can the answer be found
v~a spontaneous symmetry breaking?
ii) \lliy is the rate of ~~ two orders of magnitude higher

than K+ -7- 2n? Do we understand octet dominance?
iii) \lliy do _
m(wa)
_ and .:::..:.:.:..L
m(w ) not coincide with the observed mas-
ses vector mesons? Is this because of SU(3) breaking?
4 A. ZICHICHI
iv) Is there any link between the var~ous symmetry breaking

observed in nature:
Why is the world not as symmetric as expected if it were

simple? Is the reason spontaneous symmetry breaking?
5. DO WE UNDERSTAND LEPTONS?
i) \Vhy are the electromagnetic properties of electrons and

muons identical? Is the muon a heavy electron? If so
why is the only difference between "e" and "]1" in their
leptonic numbers, i.e. in the domain of weak interac-
tions, where nobody knows how to construct mass differen-
ces of the order of 10 2 MeV?
ii) Another heavy lepton of the standard type with its own
neutrino -- seems to show up, with standard electromag-
netic properties, at SPEAR. Why another lepton, and so
heavy,
m ~ 1.9 GeV/c 2 ?
iii) Is the role of the leptons to be a spy of the elementary

number of quarks
(~~) C)
( (:)
needed to build hadrons?

6. DO WE UNDERSTAND THE INCLUSIVE LEPTON PRODUCTION

IN HADRONIC INTERACTIONS?
i) According to the most recent results, the single lepton

yield produced in hadronic interactions seems to be con-
sistent with pair production. The only place for the
contribution from leptonic decays of the charmed par-
ticles is the very low Pt range, still open to investiga-
tion. Do we really understand the inclusive lepton/pion
ratio, allover the energy range investigated?
7. DO WE UNDERSTAND HIGH-ENERGY NEUTRINO REACTIONS?
i) The ratio of antineutrino to neutrino cross-sections, ~s
~ 1/3 at low energies, but seems to be rising at energies

above 50 GeV. What is the origin of this rising? Still
another quantum number?
ii) And what about the y-anomaly?
iii) What is the origin of the dileptons observed in high-

energy neutrino reactions?
iv) What is the origin of positron plus strange particles?

Is it charm? Are all these phenomena correlated?
v) Is it true that charm is not produced in neutral current

events?
Let us now visit the field of "standard" strong interac-

tions.
8. DO WE UNDERSTAND TOTAL, ELASTIC, AND DIFFERENTIAL
CROSS-SECTIONS?
i) Why ~s the elastic scattering imaginary at high energy?

Or, ~n Regge-Ianguage, why do even signatures dominate
and odd signatures become negligible? Is this because
diffraction is shadow scattering? Do we understand dif-
fraction scattering?
6 A. ZICHICHI
ii) Why is the Pomeron slope so small?
iii) lfuy is
o(PP)elastic I
o(PP)total
" 5
and already energy-independent?
iv) Why are meson cross-sections smaller than baryon cross-

sections? More precisely, do we understand the inequali-
ties of the total cross-sections:
0- > 0 > 0 _ > 0 + > 0 _ > 0 + > > 0 ?

pp pp TI P TI P KP K P Jp
and the identical forward slope inequalities
b- > b > b - > b + > b - > bK+p > b > b ?

pp pp TI P TI P K P p Jp
In particular, do we understand why the (J,p) slope is so
small?
v) Why do total cross-sections rise with energy as much as

they do?
9. DO WE UNDERSTAND THE HIGH-ENERGY PHENOMENA OBSERVED IN NUCLEI?
i) Why is the exponent of the A-dependence for high PT

events so high? Can this be explained in terms of co-
herent nuclear effects?
ii) What about the multipion diffractive production and its

A-dependence?
And concerning the structure of the hadrons:
10. DO WE UNDERSTAND THE INCLUSIVE ELECTROMAGNETIC

COUPLING OF THE HADRONS?
i) At relatively low energy we observe that the inclusive

electromagnetic coupling of the hadrons is point-like.
Is it because the hadrons are made of super-elementary
constituents? And if so, why do they not show up? Is
the answer: confinement?
ii) Departures from scaling have been observed in high-energy

~-induced deep inelastic processes. Are these departures
understood in terms of new quantum numbers being excited
within the nucleon constituents? Or are they within stan-
dard field theory expectation?
iii) Are the deuteron corrections really understood in the

ratio of the proton to neutron structure functions? What
is the value of
when w -+ 1 ?
If we insist on the hadron structure:
11. SHOULD WE BELIEVE IN QUARKS?
i) Hhy is SU(3) there? Is this because the so-called ele-

mentary particles are made of quarks?
ii) Hhy is 3/2 the maximum value of isospin found in the

baryon spectrum and 3 the maximum value of strangeness?
Is this because baryons are made of three quarks?
iii) Why is
a(7Tp) 2
?
a(pp) "3
Is this because mesons are made of two quarks and baryons
of three quarks?
iv) Do the electromagnetic form factors of pions and nucleons

have different q2-dependence? Do we understand the elas-
tic electromagnetic form factors of elementary particles?
v) Why do single quark transitions dominate? i.e. \fuy ~s it

so easy to have spectator quarks?
vi) Why have quarks so far not been found? Is it because

quarks are geometrical entities deprived of physical
8 A. ZICHICHI
meaning? or because the confinement theories predict the

truth? or because the production process has not been
correctly investigated in past experiments?
Here the problem is that of mastering strong interactions.

Can this ever happen?
***
Let us go into deeper problems.
12. SHOULD WE REALLY BELIEVE IN ANOTHER FUNDAMENTAL INTERACTION?
i) Why does
K -+ 2n
L
go? Or more generally, why ~s CP violated? Is this
because a new interaction -- superweak -- is at work?
13. DO WE UNDERSTAND THE UNIVERSALITY FEATURES OBSERVED IN NATURE?
i) Why ~s the weak charge universal?
ii) Why ~s the electric charge universal?
iii) Why are the proton, the neutron, the AD, as examples of
baryonic states, and the e-, ~-, v e ' as examples of lep-
tonic states, all left-handed when they interact weakly?
Is this due to the fact that they sometimes transform
from one to the other? Is this related to the fact that
the bare electric charges of the electron and proton
(i.e. of the two best measured so far) are equal?
14. SHOULD WE BELIEVE IN GAUGE THEORIES?
i) We do not observe any of the processes predicted by

theory to have infinite rates, such as all higher order
weak interaction processes and electromagnetic mass dif-
ferences. Are we sure that something physically very
relevant is not missing which causes all attempts to have
a renormalizable theory of weak interactions to fail?
Can the way out be gauge theory?
ii) All phenomena so far observed can be described in terms

of six interactions. How many fundamental interactions
are there? Are they all orthogonal to each other, or
have they a common origin?
iii) What about instantons?
15. CAN INTERNAL AND LORENTZ GROUP SYMMETRIES BE UNIFIED?
We observe regularities which go beyond SU(3); for example

SU(6):
explains the empirical agreement between the coefficients

of the Gell-Mann-Okubo mass formula for baryon octet and
decuplet, by the merging of these two SU(3) multiplets in
a 56 multiplet of SU(6). Furthermore, the ratio of mag-
netic moments ~ /~ is predicted to be -2/3 (exp. -0.68);
n p
transition magnetic moments are in qualitative agreement
as well.
explains the initial degeneracy of ~ and w.
gives the correct amount of (w-~) mixing.
predicts the equality
~ -m~ = ~* - m~
which is checked by experiment, as well as the Johnson-
Treiman relations for (TIp) and (Kp) total cross-sections,
and the relations between decay amplitudes in non-leptonic
strange particle decays.
i) Is all this because unitary spin and Dirac spins are cor-
related?
16. SHOULD WE BELIEVE IN QUANTUM FIELD THEORY?
Most of our understanding in subnuclear physics is based on

quantum field theory (QFT). The only working example of QFT is
10 A. ZICHICHI
quantum electrodynamics: everybody has failed to detect its break-

ing, even at the highest colliding (e+e-) energies.
i) Is QFT viable as a mathematical stru~ture?
ii) At a more fundamental level, is quantum mechanics

credible?
17. BEYOND PRESENT KNOWLEDGE
i) Quantum mechanics is based on complex numbers; can an

extension to quaternions and octonions lead to a deeper
understanding of particle physics?
ii) At a very fundamental level, what do we know about space?

Why is it nearly flat?
iii) The proton lifetiITle is"T 'V lei 0"T. at leas t. Why
P unlverse
is the proton so stable?
* * *
In spite of the classification into many groups~ many problems
are obviously correlated. This is again a crucial point of inves-
tigation which will be developed in the years to come. The above
series of problems~ some very hot and fashionable~ others well-
known since so long that they are even forgotten~ testify the vital
role that Subnuclear Physics has in present day scientific know-
ledge.
WHY IS THERE CHARM, STRANGENESS, COLOUR AND ALL THAT?
Harry J. Lipkin*)
Argonne National Laboratory, Illinois 60439

and
Fermi National Accelerator Laboratory, Batavia, 11.
1. INTRODUCTION
Once upon a time physicists believed that matter was made of

protons and electrons. Then the neutron was discovered. There were
now two particles, the proton and the neutron, which were very simi-
lar, yet they were also different. Now there are many particles
classified in groups containing members which are similar and also
different. Exactly how are they similar? Exactly how are they dif-
ferent? Why do particles appear in such groups? These are some of
the fundamental questions to be explored in these lectures.
The neutron and the proton have similar masses and strong in-
teractions. They have different electric charges and electromag-
netic interactions. The similarity of their strong interactions is
expressed formally by the principle of charge independence of nuclear
forces and by the symmetry of isospin invariance. The symmetry is
broken by the electromagnetic interactions which do not conserve iso-
spin but only the z component or electric charge. This symmetry
*) On leave from the Weizmann Institute of Science, Rehovot, Israel.
11
12 H. LIPKIN
breaking removes the degeneracy of the nucleon doublet, chooses the

eigenstates of electric charge as the physical particles, and intro-
duces a mass splitting between them.
The nucleon example shows the two kinds of internal quantum

numbers now used to classify particles:
1) Additive quantum numbers, conserved like charge or approximately

conserved like strangeness.
2) "Non-Abelian" quantum numbers like isospin which label families

of particles. These are associated with operators which change
the members of a given family into one another. Thus, they do
not commute with the charge operators and are called non-
Abelian.
The non-Abelian quantum numbers define families or supermu1ti-

p1ets of related particles. The additive quantum numbers label the
members of the families and distinguish between them. Such a multi-
plet structure arises naturally in any model of hadrons built from
basic building blocks in the same way that nuclei are built of nu-
cleons. The mass number and charge of a nucleus are linear combina-
tions of the number of neutrons and the number of protons in the nu-
cleus. The isospin of a nucleus is determined by the permutation
symmetry of the basic building blocks. In models where quarks are
assumed to be the basic building blocks of hadrons there are several
different types of quarks having different values for additive quan-
tum numbers.
The internal degrees of freedom which label the quantum numbers

of quarks are called flavours and colours. The values of the addi-
tive quantum numbers for any given hadron are linear combinations of
the numbers of quarks of a given flavour and colour, in the same way
that the additive quantum numbers for a nucleus are related to the
number of neutrons and protons. The non-Abelian quantum numbers are
WHY IS THERE CHARM, STRANGENESS, AND COLOUR? 13
related to permutation symmetries and the behaviour under transfor-

mations which change the colour and flavour of quarks. But unlike
nuclear physics where the nucleons are known and their properties
and quantum numbers have been measured, quarks have not been ob-
served. Thus the additive and non-Abelian quantum numbers of ha-
drons were discovered experimentally and are well established inde~
pendently of the validity of the quark model. This raises the ques-
tion: Why do hadrons have Abelian and non-Abelian quantum numbers
which suggest that they are made of quarks when quarks are not ob-
served as free particles in nature?
Some examples of the additive quantum numbers and the associated

non-Abelian symmetries are listed in Table 1.1. The question marks
indicate cases where either an additive quantum number of a non-
Abelian symmetry is known, but the companion quantum number has not
been established and it is not clear whether it exists, or is obser-
vable.
Table 1.1
Additive quantum numbers and non-Abelian symmetries
Additive Quantum Numbers Non-Abelian Symmetries
Electric charge Isospin

Strangeness SU(3) U spin
Charm SU(4)
? Colour
Baryon number ?
Lepton number ?
supm __ ~
Electron number ?
Muon number ? SpIN
Off in the corner is spin on the boundary between internal and

external degrees of freedom. Although intrinsic spin is a property
of a particle and is determined by its nature or intrinsic structure,
.j>.
Table 1.2
Symmetry algebras in physics
Laboratory
Space time Hilbert space
multiplet structure
conservation laws operator algebras
of spectrum
Rotations
Px ... Px cos e+ Py sin e 1/10. ... L Co.[3 1/1[3
[3 _J,M J n, n + i
OJ M = -j, +J
Conservation of angular -
momentum [J , J J= iJ ; [J , Jz
x y z
Isospin transformations
Charge independence of Neutrons and protons

nuclear forces [;) ... [..~.. )[;) ~have similar properties
SU(3) algebra SU(6) Isospin-strangeness

Quarks?
- 8 generators ~multiplets spin
Gauge theories unifying .----Peculiar weak currents

GIM mechanism SU (4)
weak and e. m. No nQ = 0; ns = 1
~Charmed particles
charmonium
Colour degree of freedom Baryons have wrong

Quark confinement
SU(3)colour ~ statistics
Condenser plate model ~----
Throw wrong states out of
Non-Abelian gauge theories - Quarks are not seen
Hilbert space
t
Asymptotic freedom Scaling in deep
Parton models -- free quarks inelastic scattering
:c
Baryons and fermions
Supersymmetries? ~ Graded Lie algebras r
-. eX1st ."
A
Z
it is also a physical angular momentum and can be rotated by interac-

tions in space time. Rotational invariance is a symmetry which com-
bines rotations in space time with rotations of the intrinsic spins
of the particles. In non-relativistic theory, one can postulate
symmetries in which the dynamics are invariant under separate rota-
tions of intrinsic spin and space time. The impossibility of such
a separation in a relativistic theory has led to many difficulties
in including spin together with internal symmetries 1 ) in symmetry
groups like SU(6). These difficulties are outside the scope of the
present lectures and will not be discussed further.
The transformations in the space of these internal degrees of

freedom are described by symmetry algebras. These are well known in
other areas of physics but they appear in particle physics a very
different way. This is illustrated in Table 1.2. Conventional ap-
plications begin in space time, then go to Hilbert space and then to
the laboratory2). One begins in space time with a symmetry principle
like rotational invariance which requires the equations of motion to
be invariant under certain transformations. The dynamical variables
describing the system are classified according to their behaviour
under these transformations; e.g., as scalars, vectors, tensors,
etc., under rotations. Conservation laws like conservation of angu-
lar momentum are seen to follow from these invariance principles.
The next step extends the classical implications of symmetries

to the quantum theory where states of the system are described by
vectors in Hilbert space and dynamical variables by operators. To
each symmetry transformation in space time there corresponds in
Hilbert space a linear transformation of state vectors into one
another. These state vectors can be classified into groups called
multiplets or representations of the symmetry algebra which form
closed sets transforming into one another under the symmetry. One
16 H. LIPKIN
also finds operators like the angular momentum operators which gene-
rate the symmetry transformations. The commutation relations among
these operators generate an algebra. Analysis of the algebra leads
to new operators like J2 which commute with all of the generators
and determine the structure of the mUltiplets whose states transform
into one another under the algebra.
The next step is into the laboratory to examine those conse-

quences of the symmetry algebra description in Hilbert space which
are directly verifiable in experiment. The invariance of the Hamil-
tonian under symmetry transformations means that all eigenfunctions
of the Hamiltonian connected by symmetry transformations must be de-
generate. The observed spectrum of states thus shows a multiplet
structure with states labelled by quantum numbers determined by the
symmetry algebra. In the case of rotational invariance, the multi-
plet structure consists of states labelled by quantum numbers J and
M. The non-Abelian quantum number J labels the entire multiplet
which consists of 2J + 1 states and the additive quantum number M is
the eigenvalue of the operator Z and takes on values from -J to +J
z
in steps of unity through the multiplet.
In particle physics everything goes backwards. We do not start

by an invariance principle in space time which requires invariance
under isospin transformations and end with the prediction of isospin
multiplets like the proton and the neutron. We start at the bottom
and observe a multiplet structure of the spectrum. There are two
states, the neutron and the proton, with very similar properties.
We then go back into Hilbert space and ask what are the transforma-
tions which would give rise to the observed multiplet structure. We
find the SU(2) algebra which transforms neutrons and protons into one
another. We then ask what kind of description in space time with a
Lagrangian or Hamiltonian would naturally incorporate the symmetry
that leads to invariance under these transformations in Hilbert space.
WHYISTHERECHARM.STRANGENES~ANDCOLOUR? 17
The answer in this case is a model in which all complex nuclei are
made from an elementary doublet building block, the nucleon, if the
forces which bind nucleons together to make nuclei are charge inde-
pendent.
The next multiplet structure observed in the laboratory was

groups of several isospin multiplets having different values of
strangeness and the same eigenvalues for all other conserved quantum
numbers and similar masses. The search for the right symmetry alge-
bra to describe this multiplet structure in Hilbert space took a
long time because there were no obvious elementary building blocks,
like the nucleon in nuclear physics, and there was no single obvious
candidate for the symmetry group. The correct SU(3) algebra was
eventually found and called the eight-fold way because it has eight
generators and the lowest lying baryon and meson states were classi-
fied in the octet representation, the same representation as the
generators. The search for a dynamical model which would lead to
this symmetry in Hilbert space began with the puzzle of why the sym-
metry of SU(3) should describe a system with eight basic baryons and
eight basic mesons rather than some group of transformations in an
eight dimensional vector space. One answer was that the mesons and
the baryons were not elementary objects but were composites built
from yet unknown basic building blocks with only three states. This
elementary triplet, named the quark, is very peculiar because it has
fractional electric charge and baryon number, and because it still
has not been found.
Soon after the SU(3) symmetry came SU(6) which followed from
the observation that SU(3) multiplets with different spins fit to-
gether into supermultiplets of the SU(6) algebra, as one would ex-
pect for composite models with basic building blocks having three
flavours and spin one half.
18 H. LIPKIN
The next set of symmetries were discovered neither forward nor

backward in the chain "space-time-Hilbert-space-laboratory", but via
more complicated paths. One began in space time with gauge theories
unifying weak and electromagnetic interactions 3 ) and in the labora-
tory with the observation that there were no strangeness-changing
neutral weak currents. The addition of a fourth quark flavour 4 ) and
an SU(4) symmetry with the GIM mechanisms) for suppressing neutral
strangeness-changing transitions was motivated by the peculiar struc-
ture of the weak currents observed in the laboratory and led to new
predictions that charmed particles and charmonium states should be
observed in the laboratory. The SU(4) model received a new impetus
when the neutral strangeness-conserving weak currents were observed 6 ).
There were now charged currents both strangeness-conserving and
strangeness-changing, but the neutral current conserved strangeness.
The charm model gave this kind of current in a very simple way, while
no other model gave such predictions. The completely independent
theoretical discovery that gauge theories unifying weak and electro-
magnetic interactions were renormalizable 7 ) led to a renewed interest
in these theories and the subsequent experimental discoveries of neu-
tral currents and charm gave a strong push to the development of
gauge theories.
B)
The colour degree of freed,om and colour symmetry was motivated
by three different discoveries in the laboratory. 1) That the baryon
spectrum is described simply in the quark model only if quarks have
the wrong statistics. 2) That free quarks have not been discovered
and 3) The scaling phenomenon discovered in deep inelastic lepton
scattering. In the Hilbert space, one finds that the statistics pro-
blem can be solved by introducing a new colour degree of freedom with
three colours and requiring the low-lying baryon states to be singlets
in colour SU(3). The observation that quarks, diquarks, or other
states with fractional electric charge have not been seen is explained
by pushing all states which are not colour singlets up very high in
mass or throwing them out of Hilbert space all together. The latter
is the limit of pushing them up in mass to the point where they have
infinite mass. A search for a dynamical theory described in space
time which would have these properties in Hilbert space led to non-
Abelian gauge theories which depressed all colour singlet states and
might lead to quark confinement, and the pushing up of all non-
singlet states to infinite energy. This happens exactly in a I + I
.. 9) . . . .
d~mens~onal model ,where a quark-ant~quark pa~r are l~ke a pa~r of
condenser plates and separating the plates requires infinite energy.
A different path to colour and non-Abelian gauge theories start-

ed with the observation that the experimentally observed scaling
could be obtained from parton models in which quarks behaved as free
point-like objects in deep inelastic scattering. The question of
how quarks can be so strongly bound that they can never escape, while
nevertheless behaving as quasi-free particles, led to the discovery of
asymptotic freedom 10 ) and infra-red slavery in which quarks interacted
withweak,short-range forces and strong, long-range forces. It was
then found that the only theories which had this asymptotic freedom
property were just the same non-Abelian gauge theories needed to
solve the quark statistics problem and the saturation of bound states
at the quark-antiquark and three-quark levels.
New speculations of possible additional symmetries are motiva-

ted by the existence of the additive quantum numbers of baryon number
11)
and lepton number There are suggestions that states having dif-
ferent eigenvalues of these quantum numbers could be grouped into
larger supermultiplets in a new scheme which would eventually unify
all of particle physics. The inclusion of states with even and odd
baryon number requires a different mathematical structure from the
discrete symmetries and Lie algebras used for conventional symmetries.
20 H. LIPKIN
The appropriate algebra to use in Hilbert space is called a graded

Lie algebra. These are related to dynamical symmetries in space
. 12)
t~me known as supersymmetr~es .
The unification of states having different baryon and lepton

numbers but without mixing bosons and fermions has been explored with
the aim of putting quarks and leptons, the basic building blocks of
all particles, into a unified scheme. These new speculations on su-
persymmetry and quark-lepton universality have not yet led to any
verified experimental predictions and are still at a very early stage
of development. They will not be discussed further in these lectures.
They are discussed elsewhere at this course 1 ) . Hopefully they will
lead ultimately to the answer to the question, "Why is there charm
strangeness, colour and all that?".
13)
2. STRANGENESS, CHARM AND MASS SPLITTINGS
Because symmetries are introduced backwards in particle physics,

there is no unambiguous way to introduce symmetry breaking. In con-
ventional applications like rotational invariance in atomic physics,
the symmetry is broken by a well-understood mechanism, such as an ex-
ternal magnetic field, whose transformation properties under rotations
are known. The symmetry algebra can then be used to calculate the
splittings of levels and transition matrix elements. But in particle
physics there is no underlying theory to specify the transformation
properties of the symmetry breaking interactions.
One starts in the laboratory by noting that pions and kaons have
different masses, and that additional strangeness goes with increas-
ing mass. By analogy with the breaking of rotational invariance
with a magnetic field that transforms like a vector under rotations,
one can assume that the breaking of SU(3) symmetry transforms like
the SU(3) analogue of a vector, namely an octet. This gives the
Gell-Man-Okuba mass formula. But there is no theory to tell whether

the formula applies to linear masses, quadratic masses, some exotic
power of the mass, the S-matrix, or to "reduced" matrix elements
with certain kinematic factors removed. The original folklore sug-
gested linear mass formulas for baryons and quadratic formulas for
mesons. These gave good agreement with experiment for SU(3) and
SU(6) mass formulas. But the quark model gave results which related
baryon mass splittings to meson mass splittings, in particular, the
naive assumption that the difference between strange and non-strange
quarks relates meson and baryon splittings as well as mesons and
baryons among themselves. Within the meson and baryon supermulti-
plets these quark model relations are equivalent to SU(6) relations.
But between mesons and baryons they give something new, which agrees
with experiment when linear masses are used. The situation was sum-
marized at the 1966 Berkeley conference 14 ) by the "crazy mass formula"
Q L l,Q
l<-Tt=I<~.. ~=r:~A. S ... r: (2.1)
where the L above the equality implies that linear masses should be
used and the Q above the equality implies that quadratic masses
should be used.
While there are many ways to derive some of these equalities,

no credible model includes ~ the linear and quadratic relations
involving the same vector meson mass splitting. But the experimen-
tal agreement with the crazy formula is sufficiently impressive to
suggest that it cannot be wholly accidental.
The discovery of charm allows a similar formula to be written

for the charmed states by simply replacing all strange quarks in
(2.1) by charmed quarks. The result is
22 H. LIPKIN
G L.
D -1T::: l>*-e = C:-~ (2.2)
where the last equality is left open s~nce the doubly charmed baryon
analogous to the ~ has not yet been found. This formula also agrees
with experiment, as shown in Table 2.1. Thus changing a non-strange
quark in the p to a strange or to a charmed quark produces a linear
mass shift which is equal to that produced by the corresponding change
of a quark in the ~, while the shift in squared mass is equal to that
produced by the corresponding quark change in the pion.
An interesting relation between the spin splittings of the mas-

ses of strange and non-strange baryons was given by Federman,
Rubinstein and Talmi 1s ) in 1966
(2.3)
Experimentally, the left- and right-hand sides of this relation are

307 and 294 MeV, which is rather good agreement. This relation fol-
lows from the assumption that the mass differences are due to two-
body forces which are spin dependent. The right-hand side is just
(3 k ) the difference between the interaction of two non-strange quarks
in the triplet and singlet spin states when these quarks are bound
in a non-strange baryon. The left-hand side is the same difference
for a non-strange quark pair bound in a hyperon (the particular linear
combination chosen causes the contribution from the strange quark in-
teraction to cancel out). The experimental agreement indicates that
the assumptions of two-body forces and SU(6) spin couplings in the
wave functions are good approximations.
Here again, the relation can be extended to charm by replacing

strange quarks everywhere with charmed quarks,
(2.4)
15)
Since the present experimental information on charmed baryons
gives a mass of 2260 for the Co and a mass of 2500 for a broad peak
interpreted to be the unresolved Cl - Cl* combination, it is conveni-
ent to rewrite Eq. (2.4) as
(2.5)
The left-hand side is a weighted average of the Cj and Cj* masses,

which can be roughly approximated by the value 2500 MeV for the un-
resolved peak. The left-hand side is 2456 MeV, which is in reason-
able agreement. So the spin interactions of the ordinary u and d
quarks in charmed hadrons are the same as in nucleons and hyperons.
We see that charm really behaves very much like strangeness,

and that we do not understand either!
Table 2.1
Experimental tests of crazy mass formula
a) Strangeness splittings
Q L L.Q
K - IT K* - P
~ (GeV) 0.35 0.12 0.15 0.12
t.M2 (GeV) 2 0.22 0.20
b) Charm splittings
Q L
D- IT D* - P C* - t.
~ (GeV) 1.72 1. 23 1. 26 (if MC* 2.5)

~2 (GeV) 2 3.3 3.4
24 H. LIPKIN
17)
3. HIGH ENERGY SPECTROSCOPY
3.1 Introduction
New particles are awaiting discovery with new accelerators, but

it is not clear how to look for them, particularly since the most ex-
citing new discoveries have unexpected and surprising properties.
Suggestions from theorists are of dubious value. Even when they are
right, their advice is usually useless and following it exactly usual-
ly leads to missing something crucial. But something equally crucial
can be missed by ignoring their advice. After each discovery it
usually turns out that some theorist predicted it. But dozens of
equally plausible suggestions also made at the same time led nowhere
and it was by no means obvious which approach would be fruitful.
This makes life difficult for experimentalists and program committees
trying to decide what experiments to do. But if their tasks were
easier and the outcome of experimental investigations could be pre-
dicted in advance, research would be much less exciting.
The recently discovered new charmonium spectroscopy presents an

instructive example of these difficulties. At the 1975 Palermo Con-
ference I was given credit 1B ) for predicting the discovery of these
particles on the basis of the analysis 19 ) shown in Table 3.1 of the
new particle search proposals in 1972 at Fermilab. The conclusions
were that the searches for quarks, monopoles, tachyons, etc., were
not apt to lead anywhere and that the really exciting search would
discover a particle not listed in these proposals and which the theo-
rists had not thought of. This prediction is not strictly correct
.
~f the new .
part~cles 20).
d~scovered .
s~nce Novemb er 1974 are ~n
. dee d
bound states of charmed quarks and antiquarks as they seem to be to-

day. Such states were proposed by theorists 4 ) a long time ago and
their properties were investigated in detail. However, in 1972 there
were no charm search proposals at Fermilab. Even in the summer of
::E
:I:
-<
Table 3.1 en
-i
Guide to inconclusive experiments and hypothetical articles :I:
m
:Il
m
Particle Who needs it? If not found, so what? Craziness index Signature
n
:I:
~
:Il
I. PEDAGOGICAL EXAMPLES ~
~
If MGM & YN Kills SU(3) No Good :Il
~
MO (150) -+- 5y Nobody, but why not? Nobody cares Not particularly Missing mass Z
[;)
m
II. PROPOSED SEARCHES AT NAL Z
m
en
Tachyons Nobody, but why not? Nobody cares. Try harder Very Good !fl
~
Quarks Dalitz Try harder Fair Good (fractional charge) z
c
Monopoles Dirac-Schwinger Try harder Moderately Good n
o
r
Intermediate bosons Yukawa Try harder, but credi- No Good o
bility falls C
:Il
Heavy leptons Nobody, but why not? Look elsewhere No Good ."
(spectroscopy)
Partons Bjorken-Paschos Ask Bjorken-Paschos No Good
Han-Nambu triplets Dalitz might settle Try harder Less than quarks Missing mass best
for these
Superheavy nuclei Nuclear physicists Try harder No Chemistry, not clean
III. THE REALLY EXCITING SEARCH
? Nobody has thought It will be found; it's Who knows? The theorists have not
of it there thought of it yet
t.)
til
26 H. LIPKIN
1974 when charm searches suddenly became fashionable and theorists

21)
suggested ways of looking for charm , there was no suggestion
that charmonium or hidden charm would be found long before charm it-
self or that the most fruitful search would be for very narrow states
produced in electron-posi tron annihilation. The reason why these sug-
gestions were not made is instructive. Two crucial missing links in
our understanding of hadron properties prevented the appropriate sug-
gestions from being made and taken seriously. These were the exist-
6
ence of neutral weak currents ) and the mysterious selection rule at-
tributed to Zweig, Okubo, Iizuka. and others 22 - 2s ).
In 1971 hadron spectroscopy was well described by the conven-

tional quark triplet with three quarks and no fourth quark was needed
to describe the observed states. The motivation for charm came en-
tirely from weak interactions where a number of attractive young theo
ries encountered difficulties in predicting the existence of neutral
26)
weak currents in flagrant contradiction with experiment. The in-
troduction of a fourth charmed quark with the GIM mechanismS) can-
celled out all the strangeness-changing neutral currents and removed
the disagreement with experiment. But the strangeness conserving
neutral currents were not cancelled and there was no experimental evi
dence against them, but most particle physicists assumed that this wa
simply a problem of experimental techniques. Sensitive experiments
testing strangeness-changing neutral currents were much easier than
tests of strangeness-conserving neutral currents, and there was no
obvious reason why one should be absent while the other was present.
Thus a model which looked attractive to theorists did not seem at-
tractive to experimentalists because it predicted all kinds of unob-
served experimental results and then had to introduce ad hoc cancel-
lations to get rid of them. Furthermore, the same theorists of the
Harvard group who proposed the charm model to get rid of strangeness
changing neutral currents had more complicated models 27 ) with addi-
tional heavy leptons that could get rid of all neutral currents.
There was a general proliferation of models each introducing either

new quarks, new leptons or new ad hoc couplings of electromagnetic
and weak currents. They were all equally believable and each sug-
gested different experiments to test its validity. It was hard for
an unprejudiced experimentalist to know which model should be taken
seriously or whether the whole picture of gauge theories was worth
considering seriously at al1 28 ).
Everything changed with the discovery of the weak neutral cur-

rents 6 ). It was now clear that nature had placed the strangeness-
conserving and strangeness-violating neutral.currents on a complete-
ly different basis and the most natural explanation for this dif-
ference came from the GIM mechanismS) which required the existence
of charm. So the charm model suddenly jumped from being one of many
dubious theoretical models with ad hoc assumptions not justified by
experiment to the simplest and most reasonable model available which
would explain a very striking and important new experimental result 29 ).
Attention turned to charm searches.
The charmonium states, bound states of a charmed quark-antiquark

pair were also predicted, and it was also realized that the decay of
these states would be inhibited by the same OZI selection rule which
prevents a strange quark-antiquark pair from disappearing in the ~
meson decay to produce final states without strange quarks. However,

estimates oJ the suppression factor were off by a large factor be-
cause the width of the ~ + pn decay was the only experimental evidence
available for the strength of transitions violating the selection rule.
Why the charmonium states are so much narrower is still not under-
stood.
It is now two-and-a-half years since the J particle was produced

at Brookhaven by Sam Ting and collaborators. But even though we recog-
nize the importance of Ting's discovery and great effort has gone in-
to subsequent investigations we still know very little about the pro-
duction mechanism for the J in those experiments.
28 H, LIPKIN
The OZI rule allows this J production only with an accompanying

pair of charmed particles. But there is no evidence for this charmed
,
pa1r, an d t h e J pro duct10n
' 'some
seems to go V1a h ' 25) wh'1ch
mec an1sm
violates the OZI rule.
Except for this absence of charmed pairs we know very little

about the final state in the reaction which includes the J. Thus,
it is very difficult to estimate production cross-sections for other
new objects in hadronic experiments and any extrapolation of Ting's
results for such estimates contain so many unknown factors that they
are extremely unreliable. Since the narrow width of the J is not
understood all estimates of the strength of couplings of new objects
to ordinary hadron channels are unreliable. Future experiments
might provide new insight into these fundamental uncertainties.
All properties of the charmonium states were predicted well ex-

cept for the most striking property, the very narrow width which was
crucial in their discovery. Similar theoretical considerations and
difficulties can be expected to arise in predicting the properties
of states to be discovered with new high-energy accelerators. So,
theoretical guidelines should not be dismissed but should be con-
sidered with the view that they may be even 90% correct, but a cru-
cial 10% may be missing.
3.2 Signal and noise in high-mass spectroscopy
Resonances with masses in the several GeV range have very many
open decay channels. Their branching ratios into anyone exclusive
channel are of the order of 0.1%. Since the signature for the detec-
tion of such a resonance generally picks a particular decay mode, the
signal is proportional to the branching ratio and is very small. The
crucial factor in discovering and confirming such high-mass resonances
is the signal-to-noise ratio.
It is useful to define a figure of merit F(P,T) for the produc-

tion of particle P, by observing a characteristic T of the final
state which may either be used as a trigger or as a signature for
picking out events. The trigger T may be either the full final
state like the electron pair in the decay of the J, or one of the
particles produced inclusively in the decay such as a single muon.
The figure of merit is defined by the relation
F ( P,T) = (7" ( P + X) B R (T) / 0-( T + X) (3.1)
where a(P+x) and a(T+X) denote the cross-sections inclusive for pro-
duction of the particle P and the trigger T in the reaction under
consideration and BR(T) denotes the branching ratio for the appear-
ance of the trigger T in the decay of the particle P.
Examination of Eq. (3.1) shows that the optimization of the fi-

gure of merit may best be achieved by finding a trigger T with low
inclusive production. The characteristics of the signal appearing
in the numerator will not be changed very much by choosing a differ-
ent trigger or a different production mechanism. However, the deno-
minator may be reduced by a large factor by choosing a trigger for
which the background is low. possibilities for improving F(P,T) by
reducing the noise seem to be more favourable than by enhancing the
signal. We examine three possible approaches to noise reduction.
i) Production of a low noise signal. The signal can be produced by

a mechanism which naturally has a low background, as in the production
of the ~ as a very narrow resonance in e+e- annihilation.
ii) A low noise signal signature. An exclusive decay channel can be

found which has a low production background as in the detection of the
J particle by its 1eptonic decay mode. The particular case of sig-
nature is of interest.
30 H. LIPKIN
iii) Use of background signature. Since many partial waves in the

background can appear at the high mass available and only a few in
the signal, the background may have a characteristic structure which
enables cuts in selected kinematic regions of the multiparticle phase
space to reduce the noise by a large factor.
The production of a new particle with a very low background is

possible for a narrow s-channel resonance whose cross-section is very
much enhanced over the background in a narrow energy region. This
approach can be used only for the production of resonances having the
quantum numbers available in the initial state. It is particularly
suitable for the production of vector meson resonances in electron-
positron annihilation.
For states which do not have the quantum numbers of the photon
or of the meson-baryon. nucleon-nucleon or nucleon-antinucleon sys-
tem. some possibilities exist for production via the decays of states
which do have these quantum numbers; e.g. in the production of the
positive parity charmonium states by radiative decay of the ~' and
the production of charmed particle paris by the decays of higher vec-
tor resonances.
For states not easily produced in this way and available only ~~
inclusive production there is no simple mechanism for reducing the

multiparticle background by choice of a particular production mechan-
ism. This applies to most cases of hadronic resonance production,
as in J production where no one production mechanism seems to be su-
perior by any large factor.
The triggers which have low inclusive production cross-section

in normal hadronic processes include photons and leptons produced by
electromagnetic interactions. These are suppressed by powers of a
relative to hadron production. Some examples are the lepton pairs

used as the signature for the discovery of the J particle, the pho-
tons used as the signature to discover even parity charmonium states
produced by the decay of the ~' and the two-photon and multiphoton
channels used for the possible detection of the pseudoscalar mesons.
In addition to these electromagnetic triggers which have al-

ready been used successfully, particles like the ~ and f' which are
suppressed by the OZI rule in non-strange hadron reactions might be
used successfully. These appear as signatures for states whose
branching ratios into decay channels involving ~ and ft are not sup-
pressed by significant factors over other decays. ~ signature spec-
troscopy looks attractive for states decaying into a ~ because in-
clusive ~ production without kaons is forbidden for nucleon-nucleon
and pion-nucleon reactions and the background should be small. Typi-
cal suppression factors observed experimentally for ~ production are
30)
a factor of 500 below w production in pion-nucleon reactions at
6 GeV/c or a factor of 100 below pion production at Fermilab ener-
. 31) ,. . + -
g~es The ~ ~s eas~ly detected ~n the K K decay mode at high
energies because the Q of the decay is so low that both kaons will
pass together in the same arm of a spectrometer and will not trigger
~ 31)
a Cerenkov detector set for pions An even smaller background
would be expected in ~~ spectroscopy for states expected to decay in-
to two ~'s. Examples of such states are isoscalar bosons even under
charge conjugation which have the structure of a quark-antiquark pair,
either strange, charmed, or some new heavy quark.
"Strangeonium" states of a strange quark-antiquark pair are allow-

ed by the OZI rule to decay into ~ and should have a comparatively
strong branching ratio. Such strangeonium states are of general in-
terest since no such states above the ~ or f' are well known. Our
present knowledge of charmonium spectroscopy is at present much bet-
ter than strangeonium because the low noise electromagnetic signature
32 H. LIPKIN
of lepton pairs and photons enables charmonium to be seen much more

easily. Even if spectroscopy does not lead to the discovery of
any new charmonium or "x-onium" states made from heavy quarks of
type x, the development of strangeonium spectroscopy would add to
our understanding of hadron dynamics.
The decay of charmonium or x-onium is singly forbidden by OZI

or other quark line rules and is therefore on the same footing as all
other hadronic decays which are also at least singly forbidden. Es-
timates of the branching ratios for these particles are of the
order of 0.1%. which is probably only a small factor below the pp
branching ratio. The background should be very much lower than
the pp background and therefore can provide a fruitful trigger for
such states*). The most interesting of such states at present are
the pseudoscalar states of charmonium or of the new heavier quarks
if they are there.
Single spectroscopy would be useful also in observing decays

of higher strange resonances such as K* fI. * L* :;:: * ~ * and *
which could decay into lower resonances with the same quantum num-
bers by emission above the threshold. Non-strange baryon reson-
ances at high masses have been observed by the technique of pion-
nucleon phase-shift analysis. spectroscopy may enable the disco-
very of corresponding resonances with different quantum numbers not
accessible to phase-shift analysis.
+
States like the F- meson containing both charm and strangeness
might be observed by the decay into a and a pion or lepton pair.
The TI decay mode might also be useful in the search for the exotic
four-quark states discussed in Section 6.
*) One estimate for (xx)c=+ + . is based on the analogy with ~ +

+ n which also involves annihilation of a heavy quark pair and
creation of two strange quark pairs. Another is based on the
analogy ~ + pTI and (xx) + pp and used SU(3) to relate to pp.
A partial list of states which might be detected by ~ signature

spectroscopy are:
Single ~ spectroscopy:
K * -+ K + <P (3.2a)
II. * -+ II. + ~ (3.2b)
L: * -+ L: + ~ (3.2c)
-* -+ ::: + ~
-
(3.2d)
r2 * -+ r2 + ~ (3.2e)
~ * -+K + K + ~ (3.2f)

F -+'fT + ~ (3.2g)

F -+ leptons + ~ (3.2h)
-0
Fr -+ 'fT0 + ~ (3.2i)
AS + 'fT0 + ~ (3.2j)
'fT
cp~ spectroscopy
(3.3a)
charmonium (cc)c -+ ~ + ~ (3.3b)

=+
strangeonium (s;)c=+ -+ ~ + ~ (3.3c)
x-onium (xi, where x is a new heavy quark)c=+ -+ ~ + ~ (3.3d)
Above 3 GeV the possibility of observing 3 decay arises. Vec-

tor meson states like the ~' and other higher members of the ~ family
can decay into three vector mesons. The dominant 3V final state
would be wpp but 3~ would be of the same order of magnitude in the
SU(3) symmetry limit. The 3~ state would have a unique signature and
a very low background.
The use of ~ triggers can thus lead to various kinds of interest-

ing physics. The first step is the understanding of ~ production it-
self, by examining the other particles produced along with the ~ and
34 H. LIPKIN
looking for x resonances. Understanding the mechanisms for pro-

duction can provide insight into models for particle production, even
if no new phenomena or resonances are found. But chances are that
some part of the production will be due to decays of higher resonance!
and at this stage any resonance with a decay mode is interesting.
3.2.3 ~~E~S!~~~_~iS~~!~!~~
The signal-to-noise ratio can be improved by the alternative

approach of characterizing peculiar signatures for the background
in order to enable its removal from the signal. This approach is
based on the fundamental difference between the spectroscopies of
the high mass resonances and only low-lying resonances. The con-
ventional low-lying resonances show up as peaks in cross-sections
with particular decay angular distributions against a comparatively
smooth and structureless background. At high mass the background
may have a more striking and easily identified structure than the
signal.
High mass resonances are states of low angular momentum decay-

ing primarily into multiparticle channels. Their decays reflect the
low angular momentum by containing very few partial waves all having
relatively low angular momentum. On the other hand, the background
can have very large angular momenta and a sharp structure in momen-
tum and angular distributions are present in the signal. A small
portion of the multiparticle phase space could include a very large
portion of background events. In this case the signal-to-noise ra-
tio would be improved by a cut excluding this small volume of phase
space. The exact kind of cut to be effective depends on the indivi-
dual case and could be most easily decided by examining the back-
ground and looking for its most striking features.
Consider, for example, the search for a new particle in a par-

ticular four-particle decay channel by looking for peaks in the mass
spectrum, e.g., looking for a charmed baryon decaying into A3n. The
problem is how to use the angular distributions of these four par-
ticles in the centre-of-mass system of the four-particle cluster
(hopefully the rest system of the new particle) as a means of dis-
tinguishing between signal and background. Three axes are relevant
for examining the angular distributions: (1) the direction of the
incident beam momentum, (2) the direction of the momentum of the
four-particle clusters, and (3) the normal to the production plane.
Signatures which characterize the new particle appear most clearly
in angular distributions with respect to the direction of the mo-
mentum of the four-particle cluster or with respect to the production
plane. But signatures for the noise will show up in angular distri-
butions with respect to the incident beam direction.
Background from uncorrelated particles whose mass happens ac-

cidentally to fall in the desired range should have angular distri-
butions with respect to the incident beam direction similar to those
for single-particle inclusive productions. They should be peaked in
the forward and backward directions with a rapidly falling cut-off
in transverse momentum. Background events could show forward-backward
asymmetry or a tendency to be concentrated in cones forward and back-
ward relative to the direction of the incident beam. The signal from
decay of a D meson of spin zero should show a completely isotropic
angular distribution with respect to any axis. Particles of non-zero
spin might have some anisotropy in their angular distributions if they
are polarized in production. But these will involve only low order
spherical harmonics and will not concentrate large numbers of events
in a small region of phase space. Thus a cut eliminating events in
which one or more particles appear within a narrow cone forward and/
or backward with respect to the incident beam direction could reduce
the background considerably with a negligible effect upon any signal
coming from the decay of a low angular momentum state.
36 H. LIPKIN
As an example, consider a four-particle decay into a baryon and

three pions of a state produced by a high-energy accelerator beam hit-
ting a fixed target. This state appears as a four-particle cluster
with a low mass in the several GeV region but with total laboratory mo-
mentum in the 100 GeV range. In the centre-of-mass system of the clus-
ter the momenta of the baryon and of the pions are all small and of the
same order of magnitude. In the laboratory, the baryon has a much lar-
ger momentum than the pions because of the effect of the mass on the
Lorentz transformation. If the baryon is not a proton and cannot be a
leading particle, the inclusive momentum distribution for the baryon
and the pions can be expected to be very different in the relevant
ranges. In particular, the momentum distribution for high momentum
hyperon or antihyperons could be falling rapidly in this region while
the momentum distribution for relatively low momentum pions could be
rising. This would appear in the centre-of-mass system for the multi-
particle cluster as baryons being preferentially emitted backward and
pions preferentially emitted forward. Cutting out events in which all
pions are in the forward hemisphere would thus appreciably reduce the
background, but would only remove one eighth of the signal. Using a
cone instead of a hemisphere would interfere even less with the signal
and still substantially reduce the background.
4. QUARKONIUM SPECTROSCOPY 1 7)
Among the new exciting states hopefully waiting to be discovered

are sets of positronium-like mesons made of a quark-antiquark pair
with the same flavour. These include "strangeonium" states like the
~ and f' of a strange quark-antiquark pair, charmonium states like the
J/~ family, and states made from quarks of new flavours as yet undis-
covered.
4.1 Flavour dependence of the spectrum
Strangeonium (S8) spectroscopy is still in its infancy, and is

not yet as well developed as charmonium spectroscopy, even though
strangeness was known over two decades before charm. The reason for
the comparatively slow development of strangeonium spectroscopy is
the absence of a good signature having a high figure of merit like
the electromagnetic signatures used to detect charmonium states.
The dominant decay modes of the strangeonium states are KKx which
are allowed by the OZI rule and which also appear in the background.
As a result the higher strangeonium states are expected to be broad,
have comparatively low branching ratios to electromagnetic channels,
and no striking signature different from background below the
threshold.
Charmonium (c~) has given rich experimental results because the

dominant OZI allowed decay channel, DD. is closed for a large set of
low-lying states including the radially excited s wave (the ~'), as
well as the lowest p states. Thus these states are all narrow and
have appreciable branching ratios and couplings to electromagnetic
+ - +-
channels like e e , ~ ~ , yy and yX. The vector meson states are
therefore easily produced in e+e- annihilation and photoproduction
experiments, and can also be detected by leptonic decay modes if pro-
duced by other means. Other states can be produced by cascade de-
cays of the higher vector mesons and recognized by the presence of
photons from the decay which produced them or from their own decays.
Higher x-onium states from heavier quarks with new flavours

are expected in many theoretical models, and evidence for such a state
32) 33)
has been reported E1chten and Gottfr1ed have pointed out that
such states should show an even richer spectrum than charmonium, be-
cause of theoretical arguments showing that more states lie below the
OZI allowed threshold for increasing quark mass. This threshold for
the decay of an (xx) meson is at twice the mass of the lowest (x~)
38 H. LIPKIN
state; for example, 2~ for strangeonium and 2~ for charmonium.

Eichten and Gottfried argue that the lowest vector state, analogous
to the for strangeonium and the ~ for charmonium, is farther below
the threshold as the quark mass increases, continuing the trend seen
in the and the~. Thus the range of excitation energy available
for narrow 021 forbidden resonances increases with quark mass.
4.2 Quarkonium production mechanisms
Quarkonium production for states with flavours absent in the

initial state is forbidden in strong interactions by the 021 rule.
Electromagnetic (x~) pair creation is not suppressed and is compar-
able to other (qq) production if the x-quark has an electric charge.
However, the production of (x~) from a single photon occurs only for
states with the same quantum numbers as the photon, namely odd C
vector mesons.
Processes involving the Pomeron might not be suppressed by 021.

In the SU(3) limit the Pomeron couples equally to strange and non-
strange quarks, and a factorizable Pomeron carries no information
on strangeness from one vertex to another. This is borne out by the
total cross-section for N scattering, which has no 021 suppression
factor, and is only lower than a(KN) by the same amount that a(KN)
is below a(nN). This small effect is naturally understood as SU(3)
breaking in the couplings of the Pomeron to strange and non-strange
quarks, and is not related to the connected and disconnected quark
diagrams of the 021 rule. Thus in a multiperipheral process, the ff
is emitted by a Pomeron about as easily as any other tensor meson.
34)
In the particular case of double Pomeron exchange ,one should ex-
pect to see ff production comparable to f production. In a Mueller
21)
diagram for the central region , one should also expect comparable
and w production and comparable f and ff production if the Pomeron
is approximately an SU(3) singlet as commonly believed.
There is no contradiction in the violation of OZI rule by the

Pomeron, since the connected quark diagrams used to describe Reggeon
exchanges do not apply to the Pomeron. However, in models where the
Pomeron is "built" from other trajectories 35 ), there may be some "me-
mory" of quantum numbers propagated a small distance down the multi-
peripheral chain and a consequent respect for OZI at moderate ener-
gies and low multiplicities. This question is still open. It could
be tested by looking for the f' in processes where the f is produced
by a mechanism which seems to be double Pomeron exchange, or by look-
ing at the /w ratio in the central plateau.
Experimental data on W photoproduction seem to indicate that

the coupling of the W to the Pomeron is considerably less than that
of ordinary strange and non-strange mesons. This must be taken into
account in estimating production cross-sections for new particle pro-
duction by Pomeron exchange. But this flavour dependence in Pomeron
couplings should not be confused with the OZI rule which is deter-
mined by the topological character of quark diagrams.
Hadronic production of quarkonium states may have a very dif-

ferent dependence on the spin and parity quantum numbers than electro-
magnetic production, which favours vector mesons. There are sug-
gestions that the OZI rule holds much better for vector mesons than
for pseudoscalars. In QeD, where the rule is broken by annihilation
of a quarkonium pair into gluons, three gluons are required to an-
nihilate a vector state, while a pseudoscalar can go into two gluons.
There are also experimental arguments which show that OZI violating
processes are stronger in the pseudoscalar state than in the vector
state. The absence of ideal mixing in the lowest pseudoscalar nonet
is evidence for OZI violation, since the interaction which mixes
strangeonium and non-strangeonium effectively24) violates OZI. More
recently, there is experimental evidence from radiative decays that
the OZI violating transition between charmonium states and light quark
states is stronger in the pseudos calar s tate than in the vector state 3 6).
40 H. LIPKIN
In radiative decays of charmonium to a photon and light quarks,

there are two possible transitions: (a) The photon is emitted by
the charmonium system before the transition into light quarks. In
this case the photon cannot carry away isospin and the final light
quark state must have isospin zero; (b) The photon is emitted by
the light quark system after the OZI violating transition of the
charmonium into light quarks
(cC:; Ie O)JP=~-)-+ (CCj I,::O)J P=Jflf) + l'''

(CfQj I::O J J.f= J&.Pf )+ '( (4.la)
(cC.iI=OJ.rP=~-)-+(9~j 1:0).rPI:~-}-+
(~cfj I::IfJ jP= 1f Pf )+ l (4.lb)
In case (a) the photon carries away its angular momentum and parity
before the OZI violation. and the violation occurs in a system having
the space-spin quantum numbers of the final state. In case (b) the
OZI violation occurs in a system having the space-spin quantum numbers
of the initial state before the photon carries away angular momentum
and parity. The photon can now carry away isospin zero or one. and
the final state can be both isoscalar and isovector. Thus the isospin
properties of the final state contain information on the space-spin
state in which the OZI violation occurred.
In the particular case of ~ + Py decays. the TIDy state can only

be produced by the transition (4.lb) with emission of an isovector
photon after the OZI violation has occurred in the initial vector
state. The ny and n'y states can be produced by either transition
(4.la) or (4.lb) with isoscalar photons emitted either before or after
OZI violation. Experimentally the ny and n'y decays are much stronger
o 36)
than the TI y decay ) by a factor of about 30. So, OZI violation
in the pseudoscalar state seems to be much stronger than in the vec-
tor state.
We can use this information to estimate the production of the

pseudoscalar charmonium state n in pp collisions. Assuming that
c
the difference between nc production and J production is only in
the OZI violating charmed pair creation, and that the difference be-
tween the strength of the violation in vector and pseudoscalar
states is given by the argument of radiative decays above, we ob-
tain
0" ( PP ) ... "'Ie. X A ( q~ .... c l ,j :rPc 0 - )
cr(pp). j' X A(q~ .. cc; jP=~-)
(4.2)
4.3 How to look for new quarkonium states
The charmonium experience shows that e+e- colliding beams pro-

vide a very effective means for discovering and studying the proper-
ties of vector mesons which are directly produced as s channel reso-
nances, and of other states produced by electromagnetic decays of
these vector mesons. Hadronic beams can produce these vector states,
but very little information about their properties are obtained in a
simple way because of the enormous background. If the SPEAR and DESY
results were not available to complement the information obtained
from the Brookhaven experiment, we would know very little about the
nature of the J particle, and there would be very little evidence
that it is indeed a charmonium state.
Hadronic beams might provide additional information on the pro-

perties of other states not easily seen with e +e - , such as the pseudo-
scalars. So far, the n has been seen only in one experiment at DESY
c
42 H. LIPKIN
and only in the yy decay mode. There is interest in seeing the ha-
dronic decay modes, and any ingenious method for seeing such decay
modes with hadronic production would constitute a real breakthrough
in x-onium spectroscopy. If the estimate (4.2) of the hadronic pro-
duction cross-section is reasonable, there may be some hope for de-
tecting the ne via the decay mode after production in pp colli-
sions. The figure of merit for this process can be estimated by
+ -
comparison with the detection of the J in the e e decay mode.
0" ( PP ,. "1, X ) BR (,,}, -. if If )

0" (PP .. j X) B R(J' .. e. e )
(4.3)
(pp'" ee X)

(pp .... fcpX)
Since the decay nc + is similar in nature to the decay J/~ + n,
we can assume
BR(trrc"'fCf) ~ 2. 8R (.1'/4' .... 41tf) ~ (~/35) BR (.1-+ ee) (4.4)
where we have introduced a factor 2 because only about 50% of the n

wave function, the S8 piece, contributes to the n decay mode of the
J/~, and we have substituted the experimental values for the branch-
ing ratios. Combining Eqs. (4.2). (4.3) and (4.4) then gives
(j BRC pp .... Il'}c X'" 'If X)

cr .SR(l>p,.j X .... ee X) """(30/35) ..... 1 (4.5a)
I=' ( 1fJ,,, '1Cf) cr ( pp ... e eo X)

(4.5b)
F (r ~ e.e.) "'" er( pp -+ ~ 'f X)
Thus if the background is no more than the lepton pair background,

it should be just as easy to see n
c
+ as it is to see J/~ + lepton
pairs.
31)
Results from the double-arm spectrometer experiment at Ferm~lab
showed no events, while the same run observed about 100 events of
+ -
J/~ + ~ ~. This is still consistent with the result (4.5) of equal
signal/noise and comparable signals for the two processes, because
the spectrometer had a much lower acceptance for 's than for muons.
The absence of any signal confirms that the background is low, and
that any further experiments with increased sensitivity might see a
small signal without appreciable background. Note that even three
events for at 2.8 GeV with no background would constitute serious
supporting evidence for the existence of the n , whereas several hun-
c
dred events in another decay mode against a background of thousands
of events would be ambiguous.
Similar arguments would apply to the detection of higher x-onium

pseudoscalars via the decay mode. Note that x-onium pseudoscalars
above 6 GeV would also have a ~~ decay mode which might be detectable
in a four-lepton final state.
The nA2 decay mode of the nc has also been suggested as a pos-
sible useful signature 23 ). A detailed analysis of the hadronic de-
cays of the n has been given by Quigg and Rosner 37 )
c
38)
The recent beautiful experiment at DESY reported by Schopper
. .
show~ng ev~dence for the F and F* mesons 39) is an example of how
choosing an appropriate signature minimizes background and gives
serious evidence for these particles with only a few events. The
signature in this case was three photons and a pion, with one photon
having a low energy and the other two having the mass of the n.
44 H. LIPKIN
A similar kind of signature might be used to find the nc in the

decay
(4.6)
This would give three photons, one of 300 MeV (or less if the
nc is not at 2.8 GeV but higher) and the other two having the mass
of the n, and two additional charged pions.
S. COLOUR
S.l Who needs colour?
Many reasons have been proposed for introducing colour, and not
all of them are compatible. Colour is needed by:
"0 )
1) People who like ordinary Fermi statistics for quarks and do
not like baryon models with three spin 1 quarks in symmetric rather
than in antisymmetric states.
2) .
People who l1ke .
1ntegral .charge
electr1c " 1)
3) People who believe Adler's argument for colour"Z), based on the

current-algebra-PCAC calculation of the decay TI o ~ yy. Adler's re-
sult is proportional to the sum of the squares of the charges of all
elementary fermions in the theory. The numerical experimental value
for the width of this decay agrees with predictions from a three-
colour model and disagrees with models having no colour degree of
freedom.
4) People who want to push up the ratio R =e+e o ~ hadrons/e+e- ~

+ -
~ ~ ~ , whose present experimental value exceeds the prediction from
"3 )
the simple quark model The addition of new internal degrees of
freedom pushes this ratio up, just as in TI o ~ yy.
5) People who worry about the saturation of hadrons at the quark-

antiquark and three-quark levels and want a model which explains why
states like qqq and 4qq are not found. Coloured models provide a
.. . . 8,~~,45)
natural descr~pt~on of th1s saturat~on
6) People who like non-Abelian gauge theories and quark confine-

ment 46 ). However, these people require the colour symmetry to be an
exact symmetry of nature not broken by weak or electromagnetic in-
teractions. They are unable to incorporate integrally-charged quarks
into this framework and must have fractional charges.
7) People who like to explain the ~I =~ rule by a Fierz transfor-

. ~ 7) . . .
mat~on of the four ferm10n V-A ~nteract~on.
The three-triplet model, originally suggested to allow the

three quarks in a baryon to have a symmetric wave function without
violating Fermi statistics 8 ,4o) is now called a model with "red,
white and blue" quarks. For those who find this American chauvinism
distasteful, we recommend the "Equal Opportunity Quark Model" (EOQM)
which has equal representation of black, white and yellow quarks.
For three colours and n flavours a symmetry group SU(3n) can be

defined which treats all quarks on an equal footing. This has a sub-
group SU(3)c x SU(n)f' There is no evidence for the rich hadron spec-
trum corresponding to the presence of states classified in non-trivial
representations of SU(3) The observed hadrons are assumed to belong
c
to the trivial singlet representation of SU(3) and "colour excitations"
c
of higher representations are either postulated not to exist or are
assumed to have a high mass. The colour-excited states contain ex-
actly the same coloured quarks as the observed hadrons, they differ
only in having a different permutation symmetry in the space of the
colours. This is in contrast to states with quarks of new flavours,
which can be pushed up in mass by simply postulating a higher mass
for the new flavoured charmed quarks. Colour excitations can be
46 H. LIPKIN
pushed up only by having the interaction between quarks depend on

the permutation symmetry in colour-space since different coloured
quarks all have the same mass. Interactions which confine quarks
have this property.
Models with quark confinement have an interaction between

.. . . " 8) . ..
quarks wh1ch 1ncreases w1th d1stance so that an 1nfln1te energy
is required to separate a pair. The simplest example of the con-
49)
finement is the Coulomb interaction in a 1 + 1 dimensional world .
A quark-antiquark pair behaves like a pair of condenser plates in
this world, and the force between them remains constant as they are
separated. The potential varies linearly with distance and infinite
energy is required to separate the pair. Before this happens,
enough energy is present in the field to allow a new pair to be crea-
ted. The lines of force connecting the two original quarks are bro-
ken by the new pair, and the members of the new pair couple to the
corresponding members of the old pair to make two separated bound
states with no force between them.
In three spatial dimensions, quarks are not condenser plates,

and the lines of force connecting a quark-antiquark pair can spread
out in the other two dimensions. In ordinary QED this gives the con-
ventional (l/r) Coulomb potential which does not require infinite
energy to achieve a separation. In QCD it is hoped that the non-
Abelian character bf the gauge theory produces "infra-red slavery"
which perhaps confines the lines of force to a tube and makes the
system behave like a one-dimensional system. This gives the linear
potential conventionally used for confinement. But so far there has
been no real proof that the gauge theories really predict quark con-
finement or linear potentials. The potential may have a different
form, and may not confine. A potential weaker than kr like the loga-
. h m1C
r1t . potent1a
. 1 5 0) wou ld St1. 1 1g1ve
' con f'1nement. A potent1a
. 1 wh'1Ch
requires a very large energy (e.g. hundreds of TeV) to separate

quarks would not permanently confine quarks, but would be equivalent

to confinement for experiments in the 1 TeV energy region.
In our discussions we consider the possible existence of free

quarks with a very heavy mass. This then includes the case of quark
confinement as the limit in which the free quark mass goes to infini-
ty. Note that this free quark mass is not the same as the quark mass
used in model calculations. The difference is easily seen in a model
where lines of force of the "colour field" join a quark-antiquark pair.
As the pair is separated, the lines of force cover a greater volume and
more energy is present in the field. If the quarks can actually be
separated with a finite high energy, this energy remains in the field
around the two free quarks, and covers a comparatively large volume be-
cause of the long range of the colour focre. The mass of the free quark
therefore comes from the strong long-range colour field around it.
When a quark is bound in a hadron, its colour field is confined to the
volume of the hadron and contains much less energy. Thus the mass of
the bound quark is very much less than the mass of the free quark.
In coloured quark models, the colour mayor may not be directly

observable. In models where colour is not observable, all quarks
which differ only in colour and otherwise have the same quantum n~
bers must have the same properties. In other models quarks of dif-
ferent colours have different observable properties, e.g different
electric charges. This possibility has been used to construct models
with quarks of integral electric charges. 8uch integrally-charged
coloured quarks cannot satisfy the Gell-Mann-Nishijima relation and
must have non-zero eigenvalues of a new additive quantum number which
appears in the modified Gell-Mann-Nishijima formula. The electromag-
netic current then has a component which is an 8U(3)f singlet and
which is not a singlet in 8U(3) There is a definite conflict be-
c
tween the use of integral charges and the use of colour as an exact
symmetry of nature in a non-Abelian gauge theory. If quarks of
48 H. LIPKIN
different colours have different electric charges, then the electro-

magnetic interaction breaks the colour symmetry and it is not exact.
Thus there are two incompatible approaches to colour: 1) Quark con-
finement with fractionally charged quarks; 2) Quark liberation with
integral charges. The truth might well be in between.
5.2 The deuteron world
Some insight into the coloured quark models is given by the ana-
logy of a world in which all low-lying nuclear states are made of
deuterons and have isospin zero, free nucleons have not yet been seen
and experiment has not yet attained energies higher than the deuteron
binding energy or the symmetry energy required to excite the first
I = 1 states. In this isoscalar world where all observed states have
isospin zero the isovector component of the electromagnetic current
would not be observed since it has vanishing matrix elements between
isoscalar states. The deuteron energy level spectrum (something like
that of a diatomic molecule) would indicate that the deuteron was a
two-body system, but there would be no way to distinguish between the
neutron and the proton. The deuteron would thus appear to be composed
of two identical objects which might be called nucleons. Since the
deuteron has electric charge +1, the nucleon would be assumed to have
electric charge +1. Furthermore, the nucleon would be observed to
have spin ~ and be expected to satisfy Fermi statistics. However,
the ground state of the deuteron and all other observed states would
be found to be symmetric in space and spin. Thus, the nucleon would
appear to be a spin 1 particle with fractional electric charge and
peculiar statistics.
Some daring theorists might propose the existence of a hidden

degree of freedom expressed by having nucleons of two different co-
lours. There would be a hidden SU(2) symmetry (which might be called
isospin) to transform between the two nucleon states of different
colours. All the observed low-lying states would be singlets in

this new colour (or isospin) SU(2). Since the colour singlet state
of the two-particle system is antisymmetric in the colour degree of
freedom, the Pauli principle requires the wave function to be sym-
metric in space and spin, thus solving the statistics problem.
The direct analogue of this deuteron problem in hadron quark

models is the quark model for the Q. In the conventional quark
model, the Q consists of three identical strange quarks (called
A quarks by some people and s quarks by others). with their spins
of %coupled symmetrically to spin 3/2. Since the electric charge
of the Q- is -1, the strange quark is required to have charge -1/3,
and it is also required to have peculiar statistics because the sys-
tem of three identical particles has a symmetric wave function in
all known degrees of freedom. Some daring theorists have, there-
fore, proposed the existence of a hidden degree of freedom expressed
by having strange quarks of three different colours 7 ), and a hidden
SU(3) symmetry to transform between the three strange quark states
of different colours. All the observed low-lying states are sing-
lets in this SU(3) I group. Since the colour-singlet state of
co our
the three-particle system is antisymmetric in the colour degree of
freedom, the Pauli principle requires the wave function to be sym-
metric in the other degrees of freedom, in agreement with experi-
ment and ordinary Fermi statistics. It is also possible to give
these coloured strange quarks different integral electric charges,
one with charge -1 and two neutrals, by analogue with the nucleons
in the deuteron. However, as we are concerned primarily with strong
interactions, we need not choose between models having different
electric charges for coloured quarks.
We have chosen the example of the Q for this discussion to sim-

plify the treatment of the flavour degree of freedom by considering
only strange quarks. When all flavours are considered, there are
50 H. LIPKIN
three colours for each flavour and 3n f quarks altogether. There

are two SU(n) groups, the flavour SU(n)f and the colour SU(3), which
are combined into the direct product SU(n)f x SU(3) 1 .
co our
5.3 The whys of quark model predictions
of the hadron spectrum
Let us now consider some "whys" posed by one of the outstanding

"successes" of the quark model, the prediction of the hadron spectrum.
The empirical rule that all observed hadron bound states and resonan-
ces have the quantum numbers found in the three-quark and quark-anti-
quark systems, is in remarkable agreement with experiment. Since no
alternative explanation or description has been given for this strik-
ing regularity in the hadron spectrum, this rule may constitute evi-
dence for taking quarks seriously. The quark model also predicts the
energy level spectrum of the states constructed from the three-quark
and quark-antiquark systems and observed experimentally as hadron re-
sonances. These predictions also seem to be in reasonable agreement
with experiment, but pose additional questions.
Why is the observed baryon spectrum fit only by the symmetric

51)
quark model which restr~cts the allowed states of the three-quark
system to those being totally symmetric under permutations in the
known degrees of freedom rather than totally antisymmetric, as one
expects for fermions? This can be explained by assuming that quarks
obey peculiar statistics, or that there is a hidden degree of freedom
sometimes called "colour". But this requires the additional ansatz
that all observed hadrons are colour singlets. Why and why only 3q
and qq? Why not other configurations? Why does the low-lying meson
spectrum show all the states "predicted by the quark model" without
any supplementary conditions and with no allowed states conspicuously
absent?
There is an inconsistency between the observation of bound

states in all channels for qq scattering and the absence of bound
states with quantum numbers of 2qq and 3qq. If the quark-antiquark
interaction is attractive in all possible channels, as indicated by
the presence of bound states, an antiquark should be attracted by
any composite state containing only quarks, like a diquark or a
baryon, to make a bound state with peculiar quantum numbers that
have not been observed.
In our discussion, we assume that free quarks are very heavy,

and we consider only effects on the mass scale of the quark mass.
All observed particles have zero mass on this scale. The observed
hadron spectrum is a "fine structure" which we are unable to resolve
in this approximation. This is a reasonable approach, since as long
as we are not treating spin in detail, we are unable to distinguish
between a pion and a p meson, and are neglecting mass splittings of
the order of the P-TI mass difference. We therefore are only able to
discuss whether a particle has "zero mass" and appears as an observed
hadron, or whether it has a mass of the order of the quark mass and
should not have been observed.
The question why only 3q and qq can be stated more precisely in

terms of the following three whys:
1) The triality why. With attractive interactions between quarks

and antiquarks, why are three quarks and an antiquark not bound more
strongly than a baryon or two quarks and an antiquark bound more
strongly than a meson? Note that we are not asking about four quarks
versus three quarks. Symmetry restrictions such as the Pauli prin-
ciple with coloured quarks can prevent the construction of a four-
quark state which is totally symmetric in space, spin and unitary
spin. But there is no Pauli principle which prevents an antiquark
from being added to a system of three quarks in all possible states.
52 H. LIPKIN
Thus if each quark in the baryon attracts the antiquark, some addi-
tional mechanism must be found to prevent it from being bound to the
quark system.
2) The exotics why. Even assuming some mysterious symmetry prin-

ciple which prevents fractionally charged states from being seen,
why are there no strongly bound states of zero triality, like those
of two quarks and two antiquarks or four quarks and one antiquark?
Note that we are not discussing the Rosner "baryonium" exotics which
are baryon-antibaryon resonances decoupled from the two-meson system
or Jaffe exotics bound by spin forces. We are discussing states like
an I =2 dipion resonance or bound state with a mass near the mass
of two pions. If the quarks and antiquarks in two pions attract one
another, why is there no net attraction between two positive pions
to produce a bound state or a resonance very near threshold?
3) The diquark or meson-baryon why. Why is the quark-quark inter-

action just enough weaker than the quark-antiquark interaction so
that diquarks near the meson mass are not observed, but three-quark
systems have masses comparable to those of mesons? Vector gluons
which are popular these days would bind the quark-antiquark system,
but the force they provide between identical quarks is repulsive.
Scalar or other gluons which are even under charge conjugation bind
both the quark-antiquark and diquark systems equally. If the quark
mass is very heavy, the single quark-antiquark interaction in a meson
must cancel two quark masses, while the three quark-quark interac-
tions in the baryon must cancel three quark masses. This suggests
that the quark-quark interaction is exactly half the strength of
.
t h e quark -ant1quar k"1nteract10n
"52) Such a resu 1 t can b e ach"1eve d
by a suitable mixture of vector and scalar interaction, but it is
not very satisfying to obtain such a simple fundamental property of
hadrons by a model which fits it with an adjustable parameter.
In all of this discussion, we are considering one-particle

~s, with the assumption that multiparticle states exist which
contain separated particles each having the properties we are trying
to explain. Multiparticle states pose additional problems. The al-
lowed spectrum for multiparticle states is not specified by a set of
allowed quantum numbers, but by the condition that their constituent
particles individually have allowed quantum numbers. Thus the whys
cannot be answered by general symmetry principles which apply to all
states. The triality why is not answered by a symmetry principle
forbidding all states which do not have zero triality, because multi-
particle states of zero triality must also be forbidden if they are
made of particles which individually have non-zero triality. Simi-
larly, the exotics why is not answered by a symmetry principle for-
bidding all states with exotic quantum numbers because multiparticle
exotic states made from non-exotic particles are allowed. Thus any
treatment which attempts to answer these whys must discuss both
single-particle and mutliparticle states, and must consider the
space-time properties which distinguish between them. Algebraic
arguments involving only internal symmetry groups cannot be suffi-
cient.
Our three whys involve only the strong interactions which do not
depend upon the couplings of quarks to the electromagnetic and weak
currents. The following discussion thus applies to both fractionally
charged and integrally charged models.
5.4 The coloured gluon model
We now examine the three whys. In the coloured quark descrip-

tion of hadrons, the restriction that only colour singlet states are
observed immediately solves the triality why since only states of
zero triality can be colour singlets. But requiring all low-lying
states to be colour singlets is thus equivalent to requiring all
54 H. LIPKIN
low-lying states to have zero triality; it merely replaces one ad

hoc assumption with another. What is needed is some dynamical de-
scription in which the colour singlets turn out to be the low-lying
states in a natural way. To attack this problem we return to the
fictitious deuteron world where all low-lying states are isoscalar
and which is the analogue of the coloured quark description of ha-
drons. We follow the treatment of Ref. 44.
At first this isoscalar deuteron world seems very artificial.

Why should all states with I =0 be pushed down and all states with
I .;. 0 be pushed up out of sight? But there turns out to be a very
natural nuclear interaction which creates exactly this isoscalar deu-
teron world; namely nuclear two-body forces dominated by a very
strong Yukawa interaction provided by p exchange. This interaction
is attractive for isoscalar states and repulsive for isovector states,
in both nucleon-nucleon and nucleon-antinucleon systems. It thus
binds only isoscalar states. The p exchange interaction between par-
ticles i and j can be expressed in the form
(5.1a)
-+
where t. is the isospin of par,tiele i and V contains the dependence
~
on all other degrees of freedom except isospin. If we neglect these

other degrees of freedom we can write for any n-particle system con-
taining antinucleons and nucleons,
where I is the total isospin of the system and t is the isospin of

one particle, i.e., 1 for a nucleon.
The interaction (5.lb) is seen to be repulsive for the two-body

system with I =1 and attractive for all isoscalar states. A pair
of particles bound in the I = 0 state is thus seen to behave like a
neutral atom; it does not attract additional particles. Since the
pair is "spherically symmetric" in isospace, a third particle brought
near the pair sees each of the other particles with random isospin
orientation, and its interaction with any member of the pair is de-
scribed by the average of (5.la) over a statistical mixture which is
3/4 isovector and 1 isoscalar. This average is exactly zero.
The neutral atom analogy is very appropriate for the descrip-
tion of the observed properties of hadrons. The forces between neu-
tral atoms are not exactly zero, but are much weaker than the forces
which bind the atom itself. These interatomic forces produce mole-
cules which are much more weakly bound than atoms. Similarly the
forces between hadrons do not vanish but are much weaker than the
forces which bind the hadron itself. These interhadronic forces
produce complex nuclei which are much more weakly bound than hadrons.
In the approximation where we neglect energies much smaller than the
quark mass these "molecular" effects are safely neglected.
We now generalize this picture for the coloured quark descrip-

tion of hadrons. If there are n colours, the interaction (5.1) must
be generalized from SU(2) to SU(n). The quark-antiquark system then
still saturates at one pair, but the multiquark system can be seen to
saturate at n quarks. A quark-antiquark system which is a singlet in
SU(n) exists for all values of n. However, the existence of a singlet
in the two-quark system is an accident which occurs only in SU(2) and
is not generalizable to SU(n). However, the 1=0 two-quark state is
also characterized as antisymmetric under permutation of the two par-
ticles. This antisymmetry is generalized easily to SU(n) where to-
tally antisymmetric states exist for a maximum of n particles, and the
n particle antisymmetric state is a singlet in SU(n).
56 H. LIPKIN
We now construct the analogue of the interaction (5.lb) for a

model with three triplets of different colours. Then the YUkawa
interaction produced by the exchange of an octet of "coloured gluons"
has the form analogous to (5.1). For an n-particle system contain-
ing both quarks and antiquarks,
(5.2)
where u .. depends on all the non-colour variables of particles i and

~J
j and gi (0 = 1, , 8) denote the eight generators of SU(3) 1

co our
acting on a single quark or antiquark i.
If the dependence of u .. on the individual particles i and j is

~J
neglected, the interaction energy of an n particle system can be cal-
culated by the same trick used in Eq. (5.lb) to give
(5.31'1)
where u is the expectation value of U . , integrated over the non-

~J
colour variables, C is the eigenvalue of the Casimir operator for
SU(3) 1 for the n-particle system and c = 4/3 is the eigenvalue
co our
for a single quark or antiquark. These eigenvalues are directly ana-
logous to the SU(2) Casimir operator eigenvalues 1(1 + 1) and t(t + 1)
in Eq. (5.lb).
In the approximation where all energies small compared to the

quark mass M are neglected, the interaction (5.3a) gives the mass
q
formula
The interaction (5.2) and the mass formula (S.3b) were first pro-
41) d . . .
pose d by Nambu , an the saturatLon propertLes of the LnteractLon
53)
were considered by Greenberg and Zwanziger However, the re-
markable properties of this interaction as demonstrated above in the
simplified example of the analogous deuteron world have received
little attention.
5.5 Answers to the triality and meson-baryon whys
The formula (S.3b) can test the triality why or the meson-baryon
why by showing whether observable "zero mass" hadron states exist for
a given number of quarks and antiquarks. However, it cannot test the
exotics why, since it gives no information about the spatial proper-
ties of the states. It cannot distinguish between one-particle
states and multiparticle scattering states and all zero-triality ex-
otic states are allowed as multiparticle states.
Since C is positive definite and has the eigenvalue zero only

for a singlet 45 ) in SU(3)I ,and u > 0 as is evident from the
co our -
two-body system, the state of the n-particle system with the strong-
est attractive interaction is a colour singlet. Since the interac-
tion is a linear function of n all such singlet states have zero
mass if cu/2 = M.
q
For this case
(S.3c)
The model thus gives observable hadron states for all quark and anti-
quark configurations for which C =0 states exist. Since C =0
states exist only for configurations of triality zero, this answers
the triality why.
The meson-baryon why is also answered by this interaction,

since zero mass is attained both in two-body and three-body systems.
To obtain C = 0, the two-body system must be a quark-antiquark pair,
58 H. LIPKIN
while the three-body system must be a three quark state, totally

antisymmetric in colour space. The approximation of neglecting the
dependence of u . on i and j is justified in these two cases since
l.J
there is only one pair in the two-body system, and a totally anti-
symmetric function has the same wave function for all pairs. The
va1ues~5) of the interaction parameter C-nc and the mass parameter
C/c are listed in Table 5.1 for all states of the two-body system.
These show that the quark-quark interaction in the baryon is ex-
actly half of the quark-antiquark interaction in the meson, as re-
quired for the meson-baryon puzzle. The diquark mass is thus equal
to one quark mass, since its interaction only cancels the mass of
one of the two quarks.
Table 5.1
Values of the interaction and mass parameters c-nc and c/c
System SU(3)co10ur Representation C C-nc C/c
quark-quark triplet (antisymmetric) 4/3 -4/3 1

quark-quark sextet (symme tric) 10/3 +2/3 5/2
quark-antiquark singlet 0 -8/3 0
quark-an tiquark octet 3 +1/3 9/4
The interaction averaged over all quark-quark states is seen

to be zero and similarly for all quark-antiquark states. An anti-
quark or quark added to a meson or baryon thus has a zero net inter-
action, as there can be no colour correlations between particles in
a singlet state and an external particle, and each pair feels the
average interaction over all colour states. This suggests that the
exotics puzzle is also answered, and that the states of zero mass
obtained from the interaction (5.2) for exotic quantum numbers are
mu1tipartic1e continuum states rather than bound states or resonances
5.6 The exotics why -- spatial properties

of wave functions
To examine the exotics why in more detail we consider the spa-

tial dependence of the interaction (5.2) for the specific case of
the two-quark-two-antiquark system, with an interaction u .. depend-
~J
ing only on the positions of the particles and not on momenta, spin
-+
and unitary spin. In the representation with the coordinates r. of
~
the four particles diagonal, the interactions u .. are also diagonal

~J
and can be treated as c-numbers. In this representation the inter-
action (5.2) is a 2 x 2 matrix in colour space as there are two in-
dependent couplings for four particles to a colour singlet. We
diagonalize this 2 x 2 matrix to obtain two functions of the co-
ordinates ;. which describe the spatial dependence of the interac-
~
tion in its two colour eigenstates.
It is convenient to choose a non-orthogonal basis, related by

permutations, which displays quark-antiquark couplings to C = 0,
(5.4a)
(5.4b)
where particles 1 and 2 are quarks, 3 and 4 are antiquarks and (ij)l
denotes that particles i and j are coupled to C = O. Several use-
ful identities follow from the properties of the C = 0 two-particle
state:
<O(I~> =-4/3 (5.5a)

60 H. LIPKIN
(9"cr+~3cr)lo(>=(~lcr+94G")J0(> = (5.5d)
= (~.fcr of- 94r) I~ >=(~1G"+ 930") I~) = 0
(5.5e)
<od94~~4crlf. = <~J ~4cr9aO"IQ(>= (5.5f)
- -(8/a)<0(1~>= -8/9
By operating with the interaction (5.2) on the wave functions
(5.4) and eliminating the colour variables with the aid of the iden-
tities (5.5) we obtain
and
where
Solving the secular equation for Eqs. (5.6) gives the eigenvalues
for U.
U':: - (l/ (, )(tLO( T lL/'J) - ( ., /3) lLq t

r - - - - - - - - - - - - - - (5.8)
t (.fIL) ~ f(u,oc- llf.)2. + (Ll. ot + "fJ" 2.u.q)l
If u .. is a finite range potential which vanishes at large dis-
1J
tances, the eigenvalues (5.8) reduce to those for two independent
-+
two-particle clusters for all values of the coordinates r. which cor-
1
respond to two pairs separated by a distance greater than the range
of the potential. The case Us = uq = 0 describes such a separation
between the pairs of particles (13) and (24). The corresponding
eigenvalues from Eq. (5.8) are U' -(8/3)ua, and U' = +(2/3)u a, ex-
actly those of Tables 5.1 for two separated quark-antiquark pairs
in the singlet and octet states. The case u a = = 0 describes se- Us
parated pairs of like particles (12) and (34) and has eigenvalues
U' = -(3/4)u and U' = +(2/3)u exactly those of Table 1 for two se-
q q
parated quark-quark and antiquark-antiquark systems in the triplet
and sextet states.
To test the exotics puzzle we look for coordinate configura-

tions where four-particle correlations may give stronger binding
than in two non-interacting clusters. Since ua, and Us appear sym-
metrically in (5.8). we need only consider values of Us ~ ua,' For
any value of u the value of u D < u which minimizes the interaction
a, '" - a,
(5.8) is Us = ua, with the negative sign for the quare root. This
gives
U '= - (8/3) lLex - (2./3) (tLa( - \A. q ) (5.9)
This expression is minimized by choosing the minimum value of uq

consistent with a given value of ua,' For monotonically decreasing
potentials this is achieved by placing the four particles at the cor-
ners of a square with the like particles at opposite diagonals.
62 H. LIPKIN
For a square well potential the particles can be arranged in a

square with the diagonal greater than the range of the forces and
the sides less than the range. This configuration has u = 0 and
q
forms a stable four-particle state with a binding 25% greater than
that of two quark-antiquark pairs. However, the sharp edge of the
square well is essential for this binding and does not seem reason-
able physically. For smooth potentials without sharp edges such as
Coulomb, linear, Gaussian, Yukawa or harmonic oscillator potentials
Eq. (5.9) shows that such a four-particle cluster is less strongly
bound than two non-interaction quark-antiquark pairs. and the sys-
tem simply breaks up into two clusters. This leads to a description
in which all states having exotic quantum numbers are just scatter-
ing states of particles which individually have non-exotic quantum
numbers, and answers the exotics why.
The presently accepted coloured quark model with forces from

exchange of an octet of coloured gluons provides a saturation me-
chanism in which the qq and 3q states behave like neutral
atoms 4 1 , 4 4 , 4 5 , 5 2 )D~fferent
.
parts 0
f
the b oun d state wave .
funct~on
attract and repel an external particle and the net force exactly
cancels. Thus theory and experiment now agree on the absence of
naive exotics. But the possibility exists of higher exotics. Mole-
cular-type exotics in which attraction results from spatial polari-
zation of one hadron by another have been considered, but the re-
sults (5.9) indicate that the force is insufficient to produce bind-
54)
ing. Rosner has postulated the existence of exotics from the
point of view of finite energy sum rules and duality. This approach
has been carried further by other theorists and experiments have
been suggested in a search for exotics by baryon exchange processes.
So far, there is no evidence for exotic mesons with masses be-

low 2 GeV. This has been taken as evidence against the qqqq confi-
guration for low-lying states. Although qqqq states without exotic
quantum numbers also exist, these were not taken seriously as pos-
sible configurations for the known states, because there was no good
theoretical reason why such states should be present and their exo-
tic partners should be absent. But now there seems to be evidence
that the low-lying 0++ nonet is indeed such a qqqq state~), and there
are new convincing theoretical reasons why only states with non-
exot1c quantum numb ers are seen 55)
o
5.7 A simple representation of colour couplings
For a simplified version of how normal and exotic hadrons are

constructed from coupling coloured quarks together consider simple-
minded vector coup lings in a three-dimensional colour space. In
the colour SU(3), the quark is a complex vector in three-dimensional
colour space. If we simplify this description by considering real
vectors, we use only the 0(3) subgroup of SU(3) corresponding to
real rotations and lose the distinction between quark and antiquark
which are complex conjugates of one another. But enough of the basic
physics remains to give an instructive pedagogical example. Let us
therefore consider the quark as a vector Q in a three-dimensional
colour space with red, blue, and green components denoted by
(5.10)
The colour singlet meson state is the scalar product of quark

and antiquark vectors
(S.ll)
The colour singlet baryon is the scalar product of three quark vec-
64 H. LIPKIN
Note that every quark pair in the baryon is in the antisymmetric

diquark state which is a vector product of two quark vectors
(5.13)
The antisymmetric diquark is seen to have the colour quantum num-

bers of the antiquark. The baryon can thus be written as the scalar
--
product of antisymmetric diquark vector and a quark vector
B=j)Q (5.14)
Let us now examine the states of a system containing two quarks

denoted by Ql and Q2 and two antiquarks denoted by Q3 and Q4. The
four-body system described by four vectors in the colour space can
be coupled to form a colour singlet in several ways. For example,
there is the two meson state formed by coupling the quark-antiquark
pairs separately to colour singlets
2M (5.15a)
There is also the state formed by coupling the two quarks and two
antiquarks each to an antisymmetric vector and coupling the two
vectors to a scalar
(S.15b)
This state is a possible candidate for baryonium since it could be

formed by annihilating a quark-antiquark pair in the baryon-anti-
baryon system without changing the states of the remaining quarks
and antiquarks and requiring that the state remain a colour singlet.
WHY IS THERE CHARM, STRANGENESS, AND COLOUR?
65
-
BB .... X (S .16a)
One might imagine the situation where the baryonium state X created
in some reaction would prefer to decay into the baryon-antibaryon
state via the transition (S.16) rather than to decay into two mesons
by breaking up into two quark-antiquark pairs, because the latter
transition involves changing the colour couplings. In particular,
this situation could arise if there is an appreciable spatial sepa-
ration between the diquark and the antidiquark.
One can picture lines of force joining the quarks and anti-
quarks by analogy with electrodynamics but with essential modifi-
cations following from the non-Abelian character. A colour singlet
quark-antiquark pair would have lines of force originating on the
quark and ending on the antiquark as shown in Fig. S.la. Figure
S.lb shows the two-meson system described by Eq. (S.lSa) as two
such pairs with lines of force joining the members of each pair but
no lines of force connecting the two pairs. Figure S.2 shows the
baryon described by Eq. (S.12) as three quarks at the vertices of
a triangle with lines of force between them. Here the non-Abelian
nature has new effects, with lines joining each quark and its neigh-
bour rather than quark and antiquark, and with one line acting as
a source for another, since the lines themselves carry colour. The
coupling of the baryon described in Eq. (S.14) as the product of a
diquark and a quark is seen by cutting the baryon diagram to sepa-
rate a quark from a diquark and noting that the lines of force going
from a quark to an antiquark in a meson. This again shows us that
66 H. LIPKIN
Meson Two Mesons

Fig. 5.la Fig. 5.lb
~!}
Q
Baryon
~
Diquark
Fig. 5.2 Fig. 5.3
the diquark has the same quantum numbers in colour as the antiquark.
The diquark is thus an unsaturated system with lines of force j oin-
ing the two quarks but other lines of force left out and searching
for a partner as shown in Fig. 5.3. However, the number of lines
of force originating from such a diquark are not twice the number
originating from a quark but only the same as the number originating
from a quark. Here we see again the essential difference between
non-Abelian and Abelian vector theories. The lines of force for
the electron-positron system are very much like the lines of force
for quark-antiquark system. But the lines of force for the antisym-
metrized diquark system are very different from the lines of force
in the two electron system where there are no lines joining the two
electrons and the number of lines which are unsaturated and looking
for partners is exactly twice the number from one electron.
Let us now examine the lines of force in the two configurations

(5.l5a) and (5.l5b) when the two quarks and two antiquarks are rela-
tively close together in space, but the distance between the quark
pair and the antiquark pair is much larger than the distance within
the pairs. Such a situation could be produced by annihilating a
quark-antiquark pair in an initial baryon-antibaryon state as shown
in Fig. 5.4. We see that with the baryonium colour coupling (5.l5b)
shown in Fig. 5.4b, the lines of force traversing the space between
the quarks and the antiquarks are the same as the lines of force
within a single quark-antiquark pair. However, the two meson colour
coupling (5.15) shown in Fig. 5.lb, has twice as many lines of force
traversing this space. Thus it is plausible that a baryonium-type
state X created from a baryon-antibaryon system might prefer to de-
cay by creating a quark-antiquark pair and breaking the lines of
force to return to the configuration of Fig. 5.4a rather than chang-
ing the colour couplings to the configuration (5.1Sa) shown in
Fig. S.lb which requires rearranging the lines of force to a state
68 H. LIPKIN
Vector
Q Vector Q
Baryon-Antibaryon Pair Baryonium
Fig. S.4a Fig. S.4b
of higher energy for this particular spatial configuration. This

argument is not intended to be rigorous but just to give an intui-
tive physical picture.
The two couplings (S.lSa) and (S.lSb) are not the most general
couplings to construct a scalar from four vectors. Simple analysis
shows that there are three independent couplings corresponding to
coupling any two vectors to a scalar, vector, or tensor, coupling
the other pair in the same way and coupling them both to a scalar.
However, when we return to the realistic case of complex vectors
and SU(3), there are only two independent couplings. The two
states (S.lSa) and (S.lSb) are 1inear1y.independent but not ortho-
gonal and constitute a complete non-orthogonal basis for colour
singlet state of the two-quark, two~antiquark configuration.
The state orthogonal to the two-meson state (5.1Sa), has the

two quark-antiquark pairs (13) and (24) coupled to colour octet
states rather than colour singlets and the two octets coupled to a
colour singlet. In our simplified model, with real vectors, this

includes the two states obtained by coupling the two quark-antiquark
pairs to vectors and tensors, respectively. With complex vectors,
only one linear combination of these two states is a colour singlet.
Similarly, the colour singlet state orthogonal to the baryoni-

urn state X is obtaine4 by coupling the two quarks and two antiquarks
to the symmetric sextet in SU(3) and coupling the two sextets to a
singlet. In the 0(3) subgroup of SU(3) corresponding to real vec-
tors, the symmetric sextet splits into two representations, a sca-
lar and a symmetric tensor, and scalars under real rotations can be
made either by taking the product of the two scalars or the scalar
product of the two tensors. However, only one linear combination
of these two states is a colour singlet in the SU(3) of complex vec-
tors.
6. COLOUR SPIN (MAGNETIC) EXOTICS
6.1 Introduction
The question of exotic hadron states has been confused in the

recent literature because some authors discover new things and con-
fuse the public by giving them old names like molecules which real-
ly mean something else, while others rediscover old things and con-
fuse the public by giving them new names like baryoniurn.
. ++
A more su~table analogy than a molecule for the 0 states of
two quarks and two antiquarks in the same spatial orbit is the a
particle. The question of whether or not such bound four-quark
states exist can be posed as follows: There are two analogues for
the bound quark-antiquark meson state, the deuteron and the posi-
troniurn. If the meson is like the deuteron, then two mesons should
form a bound four-quark system just as two deuterons bind together
to form a much more strongly bound a particle. If the deuteron is
70 H. LIPKIN
like positronium, the forces saturate and the residual force be-
tween the two neutral systems is very small and does not produce a
state more strongly bound than the original two-particle states.
From the experimental observation that there is no strongly bound
doubly charged state of two positive pions, we conclude that the
pion is more like positronium than like the deuteron.
However, the positronium analogy is misleading because there

is no bound state of three electrons while three quarks bind to
make a baryon. The force between two positronium atoms is nearly
zero because the repulsion between the electron pairs exactly can-
cels the attraction of the electron-positron pairs in the two posi-
tronium atoms. But in two positive pions the quark-quark force
cannot be completely repulsive because the same quarks must have at-
tractive forces to make baryons. Thus the quark-antiquark system
has many of the features of positronium. But there is an essential
new ingredient; namely non-Abelian colour and the colour-exchange
forces produced by the exchange of coloured gluons 8 ).
A red quark and a red antiquark can exchange a coloured gluon

and turn into a blue quark and a blue antiquark in the same way
that a proton and an antiproton can turn into a neutron and an anti-
neutron by exchanging charge or a charged meson. This does not oc-
cur in the Abelian case, where an electron cannot change its other
quantum numbers by emitting photons. The simplest example of a non-
Abelian interaction that we know is the model of nucleon-nucleon
and nucleon-antinucleon forces produced by pion exchange or p ex-
change. In the p exchange model of Section 5.4 the nucleon-anti-
nculeon interaction is attractive in the isoscalar state and re-
pUlsive in the isovector state. Thus there can be no bound state
of a neutron and an antineutron or of a proton and an antiproton.
The bound state is the isocolour eigenstate which is a linear com-
bination of proton-antiproton and neutron-antineutron. This means
that the proton-antiproton and neutron-antineutron states are con-

tinuous ly changing into one another by the exchange of charged p me-
sons.
This picture shows why states analogous to the hydrogen mole-

cule are not easily constructed with non-Abelian interactions. An
attraction between two hydrogen atoms can be obtained by orienting
the two states so that the proton ~n one is closer to the electron
in the other than the two protons or two electrons are to one ano-
ther. To see that this cannot be done in the model of two nucleon-
antinucleon pairs bound by p exchange let us try to put two such
nucleon-antinucleon bound pairs together so that a nucleon in one
pair is much closer to an antinucleon from the other pair than any
other pair between the two bound states. The nucleon is changing
rapidly from neutron to proton as it exchanges charged mesons with
its partner. Thus the nucleon from one pair and the antinucleon
from the other are part of the time in an isovector state where the
interaction is repulsive and part of the time in an isoscalar state
where the interaction is attractive. The net result with these iso-
spin couplings is zero interaction because the attractions and re-
pUlsions exactly cancel as shown above in Eqs. (5.1)-(5.9). Attrac-
tive forces analogous to molecular forces between atoms cannot be
obtained by introducing only spatial polarizations. Colour coupl-
ings must also be changed and the result Eq. (5.9) is a much weaker
force.
Thus the dominant forces binding quarks and antiquarks into

hadrons saturate at the qq and 3q levels. The qqqq system behaves
more like two positronium atoms than like an a particle. However,
one can ask whether residual forces much weaker than the colour
charge force might still produce binding of a-particle-like confi-
gurations. This would be analogous to having very strongly bound
deuterons which bind together into comparatively weakly bound a par-
ticles.
72 H. LIPKIN
In the early days of the quark model and SU(6) symmetry, it

was not clear whether meson states with spin greater than one would
be produced by adding more qq pairs to the qq system or by orbital
excitation of a single qq pair 56). The a-particle-like configura-
tions of two quarks and two antiquarks were considered seriously
and there were searches for states having the appropriate quantum
numbers. States with two quarks and two antiquarks all in the low-
.
est re 1 at1ve s states have J P = 0,
+ 1+ , 2+',exactly the same values
obtained for a single quark-antiquark pair in a p wave. Thus both

models predict the same angular momentum and parity quantum numbers
for the next set of excited states above the pseudoscalar and ve-
tor mesons. However, two quarks and two antiquarks can give exotic
isospin and strangeness quantum numbers not found in the qq system
with orbital excitation. After several years of searching for exo-
tics, more and more states of higher spins were found with non-
exotic quantum numbers and none were found with exotics. The or-
bital excitation model gained in favour and the a-particle configu-
.
rat10ns were f orgotten 57) .
55) .
Recently, Jaffe 1ntroduced a new 1dea He cons1ders the
binding of two quark-two antiquark states into a-particle-like con-
figurations by the spin-dependent force analogous to the magnetic
or hyperfine interaction in atomic physics. In contrast to the
atomic case where hyperfine splittings are very small compared to
orbital splittings, the hyperfine splittings in hadron spectroscopy
as indicated by the pTI and N~ splittings are of the same order of
magnitude as orbital splittings and could produce strong effects.
Jaffe finds that the lowest-lying states bound by these magnetic
interactions should appear as 0+ states with non-exotic values of
isospin and hypercharge. This natural result of the model, obtained
without any fudging or adjusting parameters, completely invalidates
the argument that the failure to find low-lying states with exotic
quantum numbers rules out a-particle-like configurations.
6.2 The flavour antisymmetry principle

55)
Jaffe has suggested the eX1stence of exot1cs bound by the
"magnetic-type" spin-dependent forces arising naturally in the
coloured-quark-gluon (QeD) models. The prediction rests on much
more general ground than the specific MIT bag model used in Jaffe's
original derivation. The essential physical input is that the N~
mass difference is much larger than the binding energy of the deu-
teron:
Mo - MN M"l+ Mp - Mel (6.1)
where n, p and d denote neutron, proton and deuteron, not quarks,

and this equation shows that there are problems of ambiguities in
both the pnA and uds notations.
The physics of Eq. (6.1) is that the dominant spin-independent

(colour charge) forces which bind quarks into hadrons saturate at
the qq and 3q states and the residual forces between colour singlet
hadrons is only of the order of 2 MeV like the deuteron binding
energy. However, the spin-dependent force responsible for the mass
difference between the N and ~ is very much larger, of the order of
300 MeV. Thus if two hadrons are brought very close together so
that the quarks in one can feel the interactions of the quarks in
the other, there is only a very weak force if the wave functions
of the individual hadrons are not changed. However, if the spins
of the quarks are recoupled to optimize the spin-dependent inter-
actions between the quarks in different hadrons, binding energies
of the order of 300 MeV are available and could give rise to bound
exotics. In the quark-antiquark system, the p-n mass splittings
show that 600 MeV is gained by changing the spins from S = 1 to
S = O.
Jaffe has simply used the N-~ and p-n mass splittings as input
for the strength of the spin-dependent interaction and calculated
74 H. LIPKIN
its effect in binding exotic configurations. Only one further in-

gredient is needed, the colour dependence of the interaction. In
colour singlet qq and 3q systems, every qq pair is in a colour
singlet state and every qq pair is in the anti symmetric colour
triplet state. Exotic configurations, even if they are over-all
colour singlets, can have some qq pairs in the colour octet state
and some qq pairs in the symmetric sextet state. The interactions
in these states are not obtainable from observed masses, and are
obtained from the spin-dependent part of the one-gluon exchange po-
tential in QeD. Evidence supporting this interaction is the agree-
ment with qualitative features of the low-lying hadron spectrum
not obtained in any other way, in particular the sign of the N-6
58)
and A-L mass splittings With this form for the interaction,
its contribution to the binding of exotic hadron states is easily
calculated by the use of algebraic techniques.
One result of the algebraic derivation is simply expressed as

the "flavour-antisymmetry principle,,59). The binding force between
two quarks of different flavours in the optimum colour and spin
state is stronger than the binding force between two quarks of the
same flavour. Although the forces are assumed to be flavour-
independent, their colour and spin dependence appear as a flavour
dependence because of the generalized Pauli principle. For maximum
binding the state should be over-all symmetric in colour and spin
together. Thus if the quarks are in the same orbit and therefore
symmetric in space, they must be flavour antisymmetric. This is
seen in the N-6 example where the I = ~ state is lower than the
I = 3/2 state even with isospin independent forces, because the
Pauli principle requires the correlation between spin and isospin
of (~, ~) and (3/2, 3/2) for a colour singlet state.
The flavour antisymmetry principle requires the most strongly

bound state of a system of quarks and antiquarks to have quarks and
antiquarks separately in the most antisymmetric flavour state
allowed by the quantum numbers. Thus, for example, the lowest

state of the six-quark system has the configuration (uuddss) with
no more than two quarks of anyone flavour.
The general question of dibaryon bound states and resonances

50)
as s~x-quark systems has been cons~dered by Jaffe ,w~th the pre-
diction of a low-lying six-quark state as a bound state or resonance
of the AA system. The exact values of the masses of these states
calculated by Jaffe can be questioned because of uncertainties in
parameters appearing ~n the bag model but certain qualitative fea-
tures are reasonably clear. The spin-dependent force between quarks
in the two baryons will be strongest in the AA system because of the
flavour-antisymmetry principle. The exact values of the masses de-
pend not only on the strength of the spin-dependent interaction,
but also on other effects not included in the model calculation and
difficult to estimate. However, if these other effects do not de-
pend strongly on flavour, dibaryon bound states, or low-lying reso-
nances are most likely to be found in the AA system.
It is interesting to note that multiquark binding lies outside

the conventional SU(6) classification of hadrons. In the SU(6)
symmetry limit the nucleon and the ~ are degenerate and the colour-
magnetic forces responsible for multiquark binding are absent. The
existence of magnetic multiquark exotics requires SU(6) symmetry
breaking, and may be related to other SU(6)-breaking effects in ad-
dition to the mass differences. One possible effect is the finite
neutron charge radius, which vanishes in the SU(6) symmetry limit.
6 1)
Carl~tz et al. have suggested that th~s results from the same
spin-dependent interaction which gives rise to the mass splittings
and have made a quantitative estimate which agrees with experiment.
It is most interesting to note that the sign of the neutron charge
radius is seen immediately from the flavour antisymmetry principle.
In the SU(6) symmetry limit the spatial separation between any
quark pair in the neutron is the same as that of any other pair
76 H. LIPKIN
and there is no spatial charge distribution. Breaking SU(6) with

the "flavour-antisynnnetric" interaction provides a stronger attrac-
tive force between quarks of different flavours and distorts the
SU(6) wave function to bring the ud pairs in the neutron closer to-
gether than the dd pair. Thus the negatively charged d quarks are
farther out on the average than the odd u quark which likes to be
closer to the differently flavoured d quarks, and the charge dis-
tribution is negative at large radius and positive at smaller ra-
dius.
So far, there is no experimental evidence for a strongly bound

M some eV1'd ence aga1nst
state, an d t h ere 1S " 1t 62) Hypernuc 1e1.
with two A's have been observed 63 ) , and are bound by only about
5 MeV more than the binding of two single A's. A AA bound state
with a much stronger binding energy would be expected to be formed
in such hypernuclei. The failure to observe this transition might
be explained by selection rules or barrier penetration factors.
But any such mechanism preventing formation of a bound state by
two A's present in the same nucleus for a time equal to the A decay
lifetime should produce even greater inhibition in any experiment
where the two A's are produced in a strong interaction collision
and are close together for a much shorter time. There may be many-
body effects in the hypernucleus which invalidate this argument;
e.g., repulsive cores in the A-nucleon interaction might prevent
two A's from coming too close together in the presence of a finite
nucleon density. But except for such effects, the existence of
the lightly bound AA hypernuclei suggests that strongly bound AA
states are not easily produced even if they exist.
For the qqqq system flavour antisynnnetry gives two very in-
55,59)
terest1ng qua11tat1ve pred1ct10ns
1) The lowest states do not have exotic quantum numbers.

2) The lowest states which have both charm and strangeness in-
clude exotics.
These predictions are simply derived by noting that a four-

body system must have two bodies with the same flavour if there
are only three flavours. Since the flavour-anti symmetry principle
requires the flavours of the quark pair and of the antiquark pair
to be different in the lowest states, the two bodies with the same
flavour must be a quark-antiquark pair. The flavour quantum num-
bers of this pair cancel one another and the quantum numbers of
the system are those of the remaining pair and therefore not exotic.
Prediction 1) gives a natural explanation for the absence of low-
lying states with exotic quantum numbers, while allowing low-lying
four-quark states with non-exotic quantum numbers. Jaffe has
called such states "cryptoexotic". Prediction 2) follows from the
observation that the flavour antisymmetry principle is easily sa-
tisfied with exotic quantum numbers when there are four flavours.
Thus exotic states with both charm and strangeness may be found in
the same mass range as the lowest F and F* mesons with both charm
and strangeness.
We now examine some experimental implications of these two

predictions.
6.3 Low-lyin g "Crypto-Exotics"
Experimental evidence seems to indicate that the lowest-lying

0++ mesons are not the quark-antiquark p-states, as formerly be-
lieved, but are, indeed, four-quark states, while the quark-anti-
++
quark 0 states are up at higher mass together with the other p-
wave excitations like the fO and A2 tensor mesons. It is signi-
ficant that the lowest states predicted by the co1oured-gluon ex-
change model form precisely a nonet of 0++ states without exotic
quantum numbers. Further experiments will tell whether these states
are indeed four-quark states and will establish the existence of
higher states.
The four-quark states constructed with flavour antisymmetry

have very different properties from the quark-antiquark states with
78 H. LIPKIN
the same quantum numbers. An isovector non-exotic, for example,

is required to have the quark constitution like (usds); it must
have a strange quark-antiquark pair to avoid having two quarks or
two antiquarks of the same flavour. Thus isovector four-quark
states will decay dominantly into modes containing strange quarks,
KK, n, etc. This is very different from the decays of conven-
tional quark-antiquark isovector states, like the A2, which decay
into non-strange channels like pn without any inhibition. This
. maske d
property ~s ~n the 0 ++ ~sovector,
. -" because ~t
the u, . ~s
. be-
low the pn and the KK thresholds and its dominant decay mode nn is
ambiguous because of mixing in the n of both strange and non-strange
components. But striking features in decay rates should be seen in
the first four-quark isovector state which is above the pn and the
KK thresholds. An unusual decay pattern ~s seen for the tensor
meson TS with the quantum numbers of the A2 but which does not de-
n
cay into pn but rather into KK, KK
* and nn and for the axial vec-
tor meson AS with the quantum numbers of the B, but with the n de-
n
cay dominant and wn forbidden. The n decay mode is particularly
interesting, since it is forbidden for all normal quark-antiquark
mesons by the OZI rule, while perfectly allowed for four-quark
states. Thus a search for n resonances might be an interesting,
way to find four-quark mesons.
The isovector non-exotic has a degenerate isoscalar companion

formed by coupling the non-strange quark-antiquark pair to isospin
zero. This isoscalar state will also decay dominantly into modes
containing strange quarks. This contrasts sharply with the be-
haviour of the degenerate isoscalar isovector doublets of the quark-
anti quark configuration like pw and fA2, where both states are
coupled more strongly to non-strange than to strange channels, and
another isoscalar state like the and the f' not degenerate with
the isovector is coupled dominantly to strange channels. Thus the
observation that the S* , the isoscalar scalar meson nearly degene-
rates with the 0, couples dominantly to kaons, and supports the
55)
classification as a four-quark state .
The observation that the 0 and the S* are scalar mesons lying
below the ~ which is the lowest vector meson in the (ss) configura-
tion has interesting implications for the new particle spectrum.
If we assume the charm-strange analogy and replace all strange
quarks in the 0, S* and ~
.the.
by charmed quarks, we pred1ct eX1st-
ence of isoscalar and isovector scalar charmonium states denoted by
oc and S* with configurations like (uu
- c~) which should lie below
c
the lowest (cc) vector meson state, namely the J/~. If we replace
*
only one strange quark in the 0, Sand by charmed quarks, we
predict the existence of exotic charmed-strange states with con-
figurations (udcs) which should lie below the lowest (cs) vector
*
meson state, namely the F. We now consider these possibilities in
more detail.
6.4 Charm-strange exotics
The four-quark states with four different flavours and the same
colour-spin couplings as the low-lying 0++ nonet constitute a set of
charmed-strange scalar mesons which are expected to lie in the same
mass range as the two-quark charmed-strange F mesons. These include
exotic states whose quantum numbers differ from those of the F by
having either the wrong sign of strangeness or the wrong isospin.
The two types of states are denoted by FI (udcs, etc. - wrong iso-
spin) and FS (udcs, etc. - wrong sign of strangeness). The "crypto-
exotic" (u~cs) four-quark state with the same quantum numbers as the
F is denoted by Fx'
-
The FI can be considered as an Fn or DK reso-
nance or bound state, the FS as a DK resonance or bound state, and
the F as an excited F coupled to the DK channel. One way to see
x
the relation of these exotics to the low-lying 0+ nonet is to note
that changing a charmed quark to a strange quark in the FI and Fx
gives a state in the 0+ nonet, while FS has no such charm-strange
analogue state. Rough estimates of their masses are near the DK
threshold. If the FS and Fx are below the DK threshold, as appears
likely, they would be stable against strong decays and decay only
weakly or electromagnetically.
80 H. LIPKIN
Table 6.1 lists these states with their quark structure, quan-
tum numbers, dominant strongly coupled channels, possible weak and
EM decay modes and their "charm-strange analogue" states in the
light quark spectrum, obtained by changing the charmed quark to a
strange quark.
One way to see how spin-dependent forces can bind a charmed-

strange exotic is by examining such an exotic configuration crea-
ted by bringing together a D+ and a K meson. The spin-dependent
force between the d antiquark in the D+ and the s quark in the K-
+
can be made stronger by recoupling the spins. In the D -K sys-
tem these two spins are completely uncorrelated, since both the D+
and K have spin 0 and are spherically symmetric. Thus the ds
system is a statistical mixture of triplet and singlet spin states,
5% S 1 like the K* , and 25% S = 0 like the K. Modifying the
wave function to give the spin coupling of the ds system a larger
S = 0 component produces additional binding on the mass scale of
the 400 MeV K-K* mass difference. A wave function 74% S = 0 and
25% S 1 instead of vice versa would gain 200 MeV in binding.
Since such recoupling of the d and s quark spins changes the spin
couplings of each of these with the other quarks, the lowest con-
figuration must minimize the total spin interaction energy of all
pairs. The dependence of the interaction on colour couplings must
also be considered, and ~s treated by the use of the SU(6) colour-
spin algebra introduced by Jaffe.
The spin-dependent interactions of the charmed quark are much

smaller than those of light quarks, as indicated by the small DD *
mass splitting relative to the pTI and KK * splittings. This is also
expected in QeD models, where the "magnetic" interaction of a quark
is inversely proportional to its mass. This also suggests that the
+ -
D K system will be bound, because recoupling the spin of the d
antiquark in a D+ to a more favourable configuration with respect
to the spins of the quark and antiquark in the K can only lose a
small amount in the more unfavourable coupling of the cd system.
The worst possible coupling can only lose the D-D * mass difference.
:;:
J:
-<
en
-I
J:
m
::0
Table 6.1 m
(")
J:
Properties of charmed-strange four-quark mesons
::0
s:
en
-I
::0
Resonance or Possible weak CS
State Quark structure (I,S,C) z
Bound state of or EM decays Analogue Gl
m
Z
+ - - - + + m
-0 en
FS csud (0,-1,+1) DK K K n n , KSKS None sn
+ + - -
- +
z
-0
FS ~sud (0,+1,-1) DK K K n n , KSKS None I o
(")
-++ + + + + + + I o
F-- cusd and c.c. (1,!,1) D-K-, F-n- n-n- o r
I o
C
-0 + - +- .::0
...,
FI (cdsu) (1,+1,+1) DOK o , F+n- n n , K K , KSKS 0-
-0 0+
FI (cdsu) (1,-1,-1) DOK o , F-n+ n+n-n o , K+K-n o
I
-+ +
F- cs(qq)I=l and c.c. (1,1,1) DKS' Fn o F-yy 00
I
+ + + +
F- cs(qq)I=O and c.c. (O,!,!) n-K F-yy, F-n S*
x S
- ~ - ~-
- ---- -- --- -------- -
~
82 H. LIPKIN
Exact mass predictions for charm-strange exotics are diffi-

cult because of uncertainties in the model. Rough estimates are
obtained by use of the charm-strange analogy, in which mass rela-
tions for systems involving strange quarks are assumed to hold
when one strange quark is replaced by a charmed quark in each state.
Examples of the success of this analogy are Eqs. (2.1)-(2.6).
While the theoretical basis of these mass relations is still not
understood, in particular why linear masses work ~n some cases and
quadratic masses work in others, the observation that whatever
works for strange quarks also works for charmed quarks suggests
that the analogy may be used to extrapolate relations from the
systems with two strange quarks to systems with one strange quark
and one charmed quark. We assume that the 6(970) is a four-quark
exotic 0+ state with the configuration (qqS8), where q denotes u
or d light quarks, and note the inequalities
M(~)+Mht)< M(cf)<2MlK) (6.2a)
Changing one strange quark to a charmed quark everywhere gives
1 ""
M(F")+M(lt) ~ M(F"IqqcS) < M(K)+M(D) (6.2b)
where the question mark expresses the uncertainty due to mixing in

the n, which is not a pure S8 state and therefore not strictly the
charm-strange analogue of the F. Thus the statement that the 6 is
below the KK threshold and decays to nn leads to the analogue that
the F should be below the DK threshold and might decay to the Fn,
but it might also be below this threshold.
We now consider the most interesting possibilities for decay

modes and signatures for the different mass ranges: Note that the
Fn decay is forbidden by isospin for strong decays of the Fx ' the
F2n decay is forbidden by angular momentum and parity for all strong
and electromagnetic decays and the F3n channel is probably well

above DK threshold.
1) All states above the DK threshold: Strong decays would be

recognized as resonances in mass plots of the DK, DK, DK and DK
systems. Decays in the Fn and F3n mode would also be allowed for
the Fr. Particularly striking signatures would be the double-
strangeness decay modes
(6.3a)
(6.3b)
2) States below the DK threshold but above Fn: The Fr would

still decay strongly to final states containing an F, but the F
x
would decay electromagnetically and the Fs weakly. The Fx + F de-
cay is a second order 0+ + 0 transition with the emission of
either two photons or no photon,
(6.4a)
(6.4b)
There is also the first-order radiative decay
(6.4c)
An intermediate F* state could be present in the decay (6.4a).

-
Another possible decay for the Fx is into the Fr , if it is
above the Fr. There would then be the cascade decay,
84 H. LIPKIN
... IV+ ... + 0 + _

FX-'" F i+(2.:~ O't e e-)'" F-+n "'{2.~ O't e e) (6.4d)
In this 0+ + 0+ transition, the e+e- decay can go via a single pho-

ton and be of second order in a like the 2y decay.
The Cabibbo favoured weak decays of the FS would be to states

of strangeness -2. States with two charged kaons would provide the
best signature for identifying these states, since neutral kaons
lose the memory of their strangeness by decaying in the ~ and KS
modes,
' " $ .... 1(- K -

F ...
1l:' +(leptons and/or pions)
+
(6.5a)
FS .... 1<+ K+ It'- ot(leptons and/or pions) (6.5b)
Strong K* signals might be expected in the KTI~ combinations, and

there should be no D present in the final state. Decays to the
four-body final states KKTITI might be the best signature, analogous
to the decays (6.3) but without the intermediate DK state and with
the possibility of one or two K*'s. Another possible signature is
in the two-body neutral decays
(6.6a)
(6.6b)
Note that the decay (6.6) can give only neutral kaons and not
charged kaons because the final states have strangeness 2 and zero
electric charge. Thus although these final states have lost the
memory of the double strangeness part of the memory remains in the
absence of the charged two-body kaon decay modes. Since nearly all
other non-leptonic decays into final states containing kaons tend
to produce equal numbers of charged and neutral kaons, the pre-

sence of anomalously large numbers of neutral kaons without the
corresponding number of charged kaons might be a good indicator
for the FS'
Because the same !inal state KSKS is produced both in the de-
cay os the FS and the FS ' there can be mixing of the two states as
in the neutral kaon system. In the approximation where CP viola-
tion is neglected, the mixing will lead to eigenstates of the mass
matrix which are CP eigenstates. The decay (6.6) will then be al-
lowed only for the eigenstate which is even under CP. The state
of odd CP will not be able to decay into two pseudoscalar mesons.
3) States below the DK and Fn thresholds but above the F. The

-+
FI would decay electromagnetically by two photon emission
The other charge states of the FI would decay weakly. The multi-
body decays would resemble the expected decay of the F with an ex-
tra pion, and the Cabibbo favoured decays would be into states of
zero strangeness. In addition, there would be the two-body decays:
r::- ++
... - +
+-
rI .... 1t'- TC (6.8a)
(6.8b)
The exotic double-charge signature for the decay (6.8a) might be a

useful indicator for this state.
The two-pseudoscalar decay modes (6.8b) are all states of even

CP for a J = 0 final state. The two states FO and F~ can be ex-
pected to mix like the neutral kaons. If CP violation is neglected
86 H. LIPKIN
then the eigenstates will be CP eigenstates and the even state will
have the decay modes (6.8b) while the odd CP state will not decay
into two pseudoscalars and will decay to three or more, in the non-
leptonic modes and into semileptonic decay modes. Note that this
P~ - F~ mixing will be much stronger than KO - KO mixing in a gauge
theory, because it can go via exchange of two intermediate W bosons,
with all vertices Cabibbo favoured and no cancellation of the GIM
type.
4) States below the F. This is highly improbable, but if the F -

is below the F, the F would now decay into the PI and the roles of
the F and FI would be reversed.
Note that if the F - FI mass difference is less than the pion

mass in either direction there will be a particle whose dominant
decay mode is electromagnetic with the emission of a low mass pho-
ton pair or electron pair. The mass spectrum of the pair will be
continuous, but its maximum must be less than the pion mass.
6.5 Are there low-lying charmonium exotics?
We have seen that the classification of low-lying 0++ mesons

as four-quark states and the charm-strange analogy relating the
qualitative systematics of systems containing charmed quarks to
known systems containing strange quarks leads to the prediction
that the charm-strange analogues of the 0 and S* denoted by 0 and
* __ c
S with the configuration (qqcc) should be lower than the J/~.
c
This suggests that perhaps the peculiar state at 2.8 GeV might not
be the pseudoscalar n at all but rather a scalar O. This would
c c
please theorists like It Hooft who want this splitting between the
pseudoscalar and vector charmonium states to be considerably less
than 300 MeV.
There is, therefore, interest in investigating the possible

existence and properties of the O. How could it be made and how
c
would it decay? Higher charmonium states might decay into or and
one or more pions, without violating the OZI rule. One might ex-
pect a11 charmonium states which can decay hadronica11y to 0 plus
c
pions to be very wide. However, angular momentum and parity selec-
tion rules seriously restrict the states which are allowed to de-
cay into n and pions. The final state 0 n has unnatural parity
c c
and even G. The I = a states produced from a strong decay of
charmonium would then have positive C. Thus of all the low-lying
+
sand p wave charmonium states, only the nc ' n~ and the X(l ) would
be allowed to decay into 8 plus a pion.
c
The final state 8
c
= 2n has I = a only if the two pions are
in an odd I = 1, odd J state like the p. The final state 0
c
+ p
has odd G and could be produced in the decay of the ~'. It would
be interesting to see whether this decay exists in present data,
whether it is ruled out and if so, what the upper limits are on
this decay. The 8 could also be produced in a radiative decay of
c
the ~.
The decay modes of the 8 include

c
Jc. ... l~ (6.9a)
dc -+ Wr (6.9b)
~c." "ITt (6.9c)
~,"ew (6.9d)
~G-+ elf (6.ge)
c>t-+ KK(KSKs> (6.9f)
de" K'" i<4c (6.9g)

88 H. LIPKIN
The possible production mechanisms for the 0c thus include
'fI' .... 1 +X(.f)~ 3.5 -. 1 +"Tr + ~c. (6.l0a)
'P'" l + de (6.lOb)
(6.lOc)
To look for the 0 , one of the decay modes (6.9) should be

c
chosen to give a convenient signature and one of the production
mechanisms (6.10) might be suitable. The decay modes (6.9c)-(6.l0f)
all have quantum numbers forbidden for the decay of the nand
c
would distinguish between the two possibilities. The decay modes
(6.9a), (6.9b) and (6.9g) are allowed for both states.
A discussion of the present experimental states of the sug-

gestion that the state at 2.8 GeV might be the n is given by
c
Gottfried 6 .. )
7. CAN WE MEASURE THE CHARGE OF A QUARK?
7.1 Quark charges, coherence and colour oscillations
The fractionally charged coloured quark model has provided an

8,52,65,66)
adequate descr1pt10n of hadron spectroscopy except for
the failure to observe the quarks experimentally. Theoretical
arguments are presently being developed to explain the unobserva-
bility of free fractionally charged particles as resulting from a
. .. 6)
fundamental conf1nement mechan1sm However, other models have
been proposed, beginning with the Han-Nambu mode1 67 ) which obtain
all the conventional results of the fractionally charged quark mo-
del from a set of integrally charged constituents. The basic dif-
ference between the fractionally charged and integrally charged mo-
del is in the description of the internal colour degree of freedom.
Fractionally charged models assume that colour is an exact symmetry

of nature, that all observable hadron states are colour singlets
and that colour is inherently unobservable. Integrally charged
models assume that colour symmetry is broken by the electromagne-
tic interaction in order to give different values of the electric
charge to quarks differing only in colour and having the same
values of all other quantum numbers. To avoid conflict with exist-
ing experimental data on observed hadron states, all such states
must be colour singlets in the integrally charged models, but the
possible existence of observable states at higher excitation is not
ruled out. Experimental observation of such colour-excited states
would establish the validity of these integrally charged models.
However, such states could lie very high in mass. The question
arises whether it is possible to distinguish between fractionally
charged models and integrally charged models below the threshold
for colour excitation 45 ,65,68).
One might think that the electric charge of the quark is ob-
servable directly from measurements of the electromagnetic coup-
lings of hadrons as shown in Fig. 7.1 since these hadron couplings
are commonly assumed to be given
by the sums of the couplings of
the constituent quarks. However,
the coupling of the electromagne-
tic current to a colour singlet
hadron depends only on the colour
averaged quark charge <Q>; i.e.,
c
the average over colour of the
charges of quarks having identical
values for all other quantum num-
Fig. 7.1 Photon absorption by bers except colour. This can be
coloured quark parton seen because a colour singlet
a) by red quark state is completely symmetric in
b) by white quark
c) by blue quark colour space and can give no
90 H. LIPKIN
information about a particular preferred direction in this space;

e.g., the charge of a red quark rather than that of a blue quark.
The colour-averaged quark charge does not contain information on
whether quarks of all colours have the same charge or whether the
charge depends upon colour. A measurement of the colour-averaged
square of the quark charge <Q2> should give this information. If
<Q2> differs from <Q>2, there must be a colour dependence of the
c c
quark charge. However, it is not simple to devise experiments on
hadrons which measure <Q2> for a given quark in a hadron, as shown
c
~n specific examples below.
One of the difficulties in interpreting results of simple

parton-type models in cases where colour symmetry is important
arises from peculiar quantum-mechanical coherence effects. As an
example, consider a model in which strong interactions exactly con-
serve colour symmetry, but electromagnetic interactions break the
symmetry by giving different electric charges to the red, white
and blue quarks. The weak interaction can also break the colour
symmetry, but might be simply expressed in a different basis from
the red, white and blue quarks which are eigenstates of electric
charge. For example, there might be purple and lavender quarks,
defined as two orthogonal linear combinations of red and blue quarks,
rotated by a Colourbibbo 69 ) angle.
Suppose that a colour singlet meson ~s given a high momentum

transfer by a strong interaction which sends the quark in the meson
to the moon, while the antiquark remains on earth. Since strong in-
teractions conserve colour, the system is still in a colour singlet
state. If an astronaut on the moon measures the electric charge of
the quark and finds that it is red, then the antiquark on the earth
must also be red, and similarly for blue or white. But if the astro-
naut does a weak interaction experiment and finds that the quark
was purple or lavender, then the antiquark on the earth must also
be purple or lavender. Thus, whether the density matrix describing
the antiquark on earth 1S diagonal in the red-blue or purple-

lavender basis depends upon whether the astronaut on the moon
chooses to measure an electromagnetic or a weak property of the
quark on the moon. This is a manifestation of the famous paradox
of Einstein, Podolsky and Rosen.
Such coherence properties arise 1n parton models where non-

Abelian symmetries are present. One generally draws diagrams like
those of Fig. 7.1 in which one parton absorbs a photon and behaves
as if it were free during the interaction. However, it is only as
free as the quark on the moon in the previous example. If it
comes from a colour singlet hadron state, a measurement of its
electric charge by a photon absorption process as in Fig. 7.1 af-
fects the properties of the rest of the system, even though there
is no interaction. This effect appears in the relative phases of
the contributions of the three diagrams shown in Fig. 7.1, where
the photon is absorbed by a red, white and blue quark, respectively.
If this phase information is ignored and the contributions of the
diagrams are added incoherently, the results obtained can have seri-
ous errors.
An example of the importance of relative phases in cases where

internal symmetries are present has been pointed out in the deep
inelastic production of exclusive final states on a pion target by
65)
the isoyector component of the photon . Since the initial state
has odd G parity, only final states with odd numbers of pions can be
produced. However, a parton model in which individual quark partons
absorb the photon and the amplitudes are added incoherently loses
the G parity information and gives equal production of states with
even and odd numbers of pions. The G parity information is con-
tained in the relative phase of the contributions from pairs of dia-
grams which go into one another under the G transformation; i.e.,
those in which the current is absorbed by a quark and by its G con-
jugate antiquark.
92 H. LIPKIN
An important effect which must be understood to avoid pitfalls

in intuitive treatments of colour is the phenomenon of colour oscil-
lations. These are analogous to the strangeness oscillations 1n
the neutral kaon system and the neutrino oscillations which have
been suggested as possibly occurring if two neutrinos have differ-
ent masses. Consider a coloured quark-antiquark state of a red
quark and a red antiquark. This is not a stationary state but
oscillates between red-antired, white-antiwhite and blue-antiblue,
just as the KO is not a stationary state but oscillates between KO
and KO as it decays and Ve and V~ are not stationary states and
oscillate in some models. The frequency of the colour oscillation
is determined, as in the neutral kaon and neutrino cases, by the
mass difference between the true stationary states of the system.
In the colour case this mass difference is not very small, like the
~ - KS or neutrino mass difference, but is very large. It is the
mass difference between the observed colour singlet mesons and the
as yet unobserved colour octet states. Thus colour oscillations
occur at a very rapid rate. In models where quarks are permanently
confined, this mass difference is infinite and colour oscillations
occur with infinite frequency. The colour of a quark is thus un-
observable, and all properties of quarks measured in confined sys-
tems must be colour-independent.
In models where quarks are not permanently confined and states

which are not colour singlets exist, the threshold for colour-octet
excitation defines a critical mass and a critical time at which
drastic changes in particle physics can be expected. For times
long compared with this time scale, the colour oscillations are so
rapid that all quark properties measured are averaged over colour
and there is no hope of distinguishing between fractionally charged
and integrally charged models. To observe the difference in pro-
perties between quarks of different colours, an experiment must
have a built-in time-scale which is short in comparison with the
colour oscillations and which can measure the charge of a quark
before it changes. This suggests that the experiment must have an

energy or mass scale which is above the colour threshold. Thus the
question of distinguishing between models below the colour threshold
becomes one of measuring short-time behaviour with lower energies.
With these difficulties in mind, we examine the possibility

of observing a colour dependence of the electric charge.
Models like the Han-Nambu model 67 ) are constructed to make

colour averages of all matrix elements of the electromagnetic cur-
rent exactly equal to those of the coloured fractionally charged
model. The difference between the currents of the two models has
no colour singlet component and its colour average vanishes. Let
us write
(7.la)
where J G is the electromagnetic current in Greenberg's coloured

'+ 0) . . .
quark model w~th the fract~onal charges of the Gel I-Mann - Zwe~g
quark model, and the arguments 8 f ,lc denote that this current trans-
forms under flavour and colour like an octet and singlet, respec-
tively. In the Han-Nambu model, ~J is a flavour singlet and a co-
lour octet. Thus
(7.lb)
The matrix elements of J thus all vanish between colour singlet

states, and J is unobservable in any measurement described by such
8,52,65)
a matrix element .
The electric charges of the quarks have the same structure as

the current operators. Thus the charge of a quark of flavour f and
colour c in the Han-Nambu model is given by
(7.lc)
94 H. LIPKIN
where QG depends only on flavour and is independent of colour, and

~Q depends only on colour. We thus obtain for the colour averages
of Q and Q2,
(7.ld)
where the value 2/9 is obtained by substituting the numerical

values of QHN(f,c).
As long as ~J has no observable effects, it is impossible to

distinguish between the integrally and fractionally charged models.
There are two possible approaches to the observation of ~J: 1) by
observing states which are not colour singlets, 2) by observing
matrix elements of operators which are quadratic in J between
em
colour singlet states. Since states which are not colour singlets
have a presumably high excitation threshold to explain the failure
to observe them to date, we consider the possibility of detecting
~J below threshold by measurements on colour singlet states of
operators quadratic in ~J.
7.1 Two-photon decays of pseudoscalar and tensor mesons
One important case where effects of colour have been observed

in a second-order electromagnetic transition is in the decay nO +
+ yy. Decays of this type of meson into two photons are assumed to
. , 32)
be descr~bed by a tr~angle d~agram . We cons~der all poss~ble
decays of common mesons which have allowed two-photon decays,

namely the pseudoscalar and tensor mesons:
(7.2a)
(7.2b)
, . -+ 2.r (7.2c)
fO -.21 (7.2d)
(7.2e)
(7.2f)
The contribution of the tria~gle diagram for each of these de-

cays is obtained by summing the diagrams for different quark fla-
vours with the appropriate weighting factors for each meson. The
transition matrix element for this diagram with a quark of flavour
f is proportional to the square of the quark charge and is given by
(7.3a)
where M is the reduced transition matrix element which contains the

dependence on all degrees of freedom except colour. Substituting
from Eq. (7.le) into Eq. (7.3a), we obtain the relation between the
transition matrix elements in the Greenberg model and the Han-
Nambu model, denoted by MG(f) and ~~N(f):
(7.3b)
(7.3c)
The expressions (7.3) can be written for the specific cases of

u,d and s quarks in the following convenient form,
96 H. LIPKIN
M(~) = 3 M (4+2.1<) /9 (7.4a)
N(d) =M(s) =-./3 M(~+2K)/9 (7.4b)
where K is a parameter describing the deviation from the fraction-

ally charged coloured quark model and we have set N = 3 x MG(f)
c
is given by setting x = 0 in Eqs (7.4) and ~N(f) is given by set-
ting K = 1. Intermediate values of K are also of interest as will
be shown below.
For the decay of a TI o which is a coherent linear combination

of a uu and dd state with equal magnitude and negative phase, the
transition matrix element is proportional to the difference MT(u) -
MT(d). This difference is seen to have the same value in both mo-
dels:
MT (u.)" - NT CeL) Go =of IV! =M T( U.)HN - MT Cd.) HN (7. Sa)
A similar equality holds for the decay of the ns which is the eighth
component of an octet and depends upon the linear combination
The statistical factor IN

in Eq. (3) has been used as evidence
in favour of colour in the experimental value of the TI o + yy decay
rate. However, Eqs. (7.5) show that it is impossible to distinguish
between fractionally and integrally charged models with this decay
or the decay of the isoscalar unitary octet meson. This is also
evident from the form of ~J in Eq. (7.lb), which is a flavour sing-
let. Squaring ~J gives an operator which has a colour singlet
component and is observable in the space of colour singlet states.

But because it is a flavour singlet, it cannot contribute to the
decay of a flavour octet state.
For the decay of a flavour singlet meson, the two models give
different results:
(7.6)
Unfortunately, this difference is not easily checked experimentally.

The physical n' meson is a mixture of singlet and octet and is at
such a high mass that the PCAC derivation for the absolute rate of
the nO decay is unreliable. Kinematic factors resulting from the
n-n' mass difference confuse any comparison of the two rates.
Thus, although P + yy decays appeared to have matrix elements

quadratic in 6J which would distinguish between the two models,
this is not feasible in practice. The situation looks somewhat
better for an ideally mixed nonet, like the tensor mesons, where
the mass degeneracy between non-strange isoscalar and isovector
states causes all kinematic factors to drop out in the ratio of the
decay rates. For the ideally mixed fO and f' decays:
M(fO) =(~/a)[M(u.)+M(dJJ (7.7 a)
M (Al) :: ('fIJI) [M(u.) -M Cel)] (7.7b)
M( f') = M(s) (7.7c)
Substituting Eqs. (7.4) into Eqs. (7.7) gives
M(f-)/M(AZ)/M(f') :: (7.8)
=[(5/3)+(4/3)t<] I~I [C.JI/3)+(2fI/3)1<)

98 H. LIPKIN
Since the decay rates are proportional to the squares of the matrix
elements, the ratio of the fO to A2 decay rates is predicted to be
9 in the Han-Narnbu model in comparison with 25/9 in the fractional-
ly charged model. Furthermore, the f' decay rate is predicted to
be larger than the A2 decay rate by a factor of two in the Han-
Narnbu model and lower by a factor of 2/9 in the fractionally
charged model. These appear to be large observable effects.
Additional possibilities of observing the differences between

the matrix elements (7.7a) and (7.7b) arise in coherent production
o . 66)
o f the f and A2 resonances by two photons ~n the react~on
(7.9)
where MO denotes a neutral non-strange meson which is a coherent

linear combination of fO and A2 and PP denotes a state of two pseu-
doscalar mesons. In the approximation where the fO and A2 are de-
generate, SU(3) symmetry and the OZI Rule give the following re-
sults for the relative cross-sections for the production of dif-
ferent PP states:
M(K+f(-) IM(KOKO)/M(~i'-rr.-)-= (7.10)

; (4+2,,) I (~+ll<)/(5+L.,.J()
where the transition matrices MT must be squared and multiplied by
appropriate kinematic factors to obtain the observed cross-sections.
The kinematic factors should be identical for the charged and neu-
tral kaon final states but may be somewhat different for the two-
pion state. The results (7.10) are easily obtained by observing
that the fO and A2 are linear combinations of the u~ and dd states
and that charged kaon pairs are produced only via the uu state, neu-
tral kaons only via the dd state and pion pairs only via the even
G fO state.
The relations (7.10) are derived under the assumption that the
fO and A2 are degenerate and have the same width. Calculations of
the charged and neutral kaon pair mass spectra show that the quali-
tative features of Eq. (7.10) remain when the masses and widths of
the physical particles are introduced and each resonance decay is
parametrized by a Breit-Wigner curve. In addition, a strong inter-
ference effect appears in the region between the ff and the A2 from
the overlapping of the tails of the resonances. With the Han-Nambu
model these interference effects should be somewhat different, and
might be used to distinguish between the two models.
Relations (7.8) and (7.10) look very promising for distinguish-

ing between the two models if the simple triangle diagram describes
the decay and its transition matrix element is given by Eq. (7.3).
However, there are doubts about the validity of this description
for the tensor mesons.
Suppose the triangle diagram of Fig. 7.2 is interpreted as the

successive emission of two photons. Then an intermediate state
exists of a quark-antiquark pair with the quantum numbers of the
photon, a vector-meson state which
is either a colour singlet and
flavour octet or flavour singlet
and colour octet. The transition
q
matrix element computed from this
diagram must include a propagator
for the intermediate state. From
Eqs. (7.1) it is apparent that J G
appears in diagrams with colour
Fig. 7.2 Triangle diagram singlet intermediate states and ~J
for two-photon de-
appears in diagrams with colour oc-
cay of a meson
tet intermediate states. If colour
octet states have a high threshold, diagrams with colour octet inter-
mediate states will be suppressed by propagators relative to diagrams
with colour singlet intermediate states. Thus the effect of the
100 H. LIPKIN
colour threshold will reduce the contribution of the terms depend-

ing upon 6J below the values given by Eqs. (7.3), (7.4), (7.6) and
(7.8). Since the contribution of J is seen from Eq. (7.3b) to be
positive definite and to be given in Eqs. (7.4), (7.6) and (7.8) by
the term proportional to K, the reduction of the contributions from
6J are expressed quantitatively by reducing the value of K from
unity in these relations.
For the case of the pseudoscalar meson decays, the existence

of the axial anomaly allows the transition matrix element to be ex-
pressed by a triangle diagram dominated by high momenta where co-
lour thresholds are hopefully no longer important. For other cases
where there is no anomaly, there is no reason to expect this domi-
nance by high momenta in intermediate states and colour threshold
effects can be important. Unfortunately, the large deviation from
ideal mixing makes the use of pseudoscalar decay rates difficult
for distinguishing between the two models. The ideal mixing of the
tensor nonet gives simple predictions, but these may be rendered
useless by colour threshold effects.
The effects in Eqs. (7.8) and (7.10) are so large that they
may still be observable even with an appreciable reduction from the
propagators of the colour octet states. The parameter K will have
a value (m1/ms)2, where m1 and ms are the masses of the colour sin-
glet and colour octet intermediate states which are dominant in the
transitions. If K is between 10- 1 and 10- 2 , there may still be a
possibility of observing these effects. For example, if K = 2%,
there will be an 8% increase in the ratio of the two-photon decay
widths of the ff and A2, a 3% increase in the ratio of the widths
of the f and A2, and a 3% decrease in the production ratio of
charged to neutral kaon pairs over the predictions of the fraction-
ally charged quark model. Thus even if the effects are small, they
appear as uniquely related discrepancies from the predictions of the
fractionally charged model in three different ways.
7.3 Deep inelastic processes
Another type of process where colour effects might be observed

is deep inelastic scattering described by the quark parton mode1 68 ).
Here again, the simplest processes cannot distinguish between inte-
grally and fractionally charged models because they are described
by matrix elements of the current between colour singlet states.
We consider the possibilities of observing a colour dependence of
the electric charge in two ways: 1) above the colour excitation
threshold; 2) by transitions depending quadratically on the elec-
tromagnetic current.
Once states which are not colour singlets are produced opera-
tors which are not colour singlets become observable. However, the
parton sum rules do not necessarily hold immediately in this new
domain with the integral quark charges of the Han-Nambu model. The
basic incompatibility between naive parton models and non-Abelian
internal symmetry must be carefully considered before drawing con-
clusions. We now spell out this incompatibility explicitly and
show the necessary conditions for validity of the naive parton mo-
del.
Consider the parton model description of the absorption of a

current by a hadron. In a model with three coloured quarks de-
noted byR, Wand B for red, white and blue, the current can be
absorbed by a quark parton of any colour. There are three contri-
buting diagrams in which exactly the same transition occurs on a
red, white or blue quark as shown in Fig. 7.1. To calculate a par-
tial or total cross-section, the transition amplitudes for the
three processes shown in Fig. 7.1 must be added coherently and then
squared. The naive parton model neglects the interference between
different diagrams and adds them incoherently under the assumption
that interference terms have random phases and average out. This
is the source of error in calculations for processes invariant under
non-Abelian internal symmetries. The symmetry imposes conditions on
102 H. LIPKIN
the relative phases of different amplitudes so that the contribu-

tion of interference terms does not average out but 1S of the same
order of magnitude as the direct terms. This has been pointed out
in the G parity example discussed above 6s ).
We now demonstrate this interference effect for colour sym-

metry in a simple example. Let us assume that the transition ma-
trix element for the absorption of a current by a quark is propor-
tional to the charge of the quark denoted by gR' gw and gB' re-
spectively. The transition amplitude between a given initial and
final state is then given by
<f IT/ ~) = 9R <tlARI i> + ~W <f'A W 'i>"'98 <fl AS' i. > (7.11)
where ~,~ and AB are reduced transition amplitudes for the red,
white and blue quark transitions with the quark charge factored out.
The values of these reduced matrix elements for different colours
are related by the colour symmetry and depend on the colour quantum
numbers of the states Ii> and If>.
For the case where both Ii> and If> are colour singlets
(7.l2a)
The contributions of the three diagrams of Fig. 7.1 are all equal
and have the same phase and the invariant amplitude A is defined
for convenience with the normalization indicated.
For a transition between a colour singlet initial state and

colour octet final states, two linearly independent colour octet
states occur. For convenience we choose as basic states, denoted
by If3> and Ifa>, the vector and scalar states under the SU(2) sub-
group of the colour SU(3) which acts only in the space of white and
blue quarks. For these transitions we find:
WHYISTHERECHARM.STRANGENES~ANDCOLOUR? 103
(7.12b)
<f 3' A w Ii. i} = - <f 3' AB' i >:: cia

of (7.12c)
i <f,I ARo'l ~ > = - <f " Aw I i 4> = (7.12d)

'= _(fvIA.B'i4>:: clf6
where the SU(3) colour symmetry relates the two transitions (7.12c)
and (7.12d), and the invariant amplitude C is normalized for con-
venience.
Substituting Eqs. (7.12) in Eqs. (7.11) we obtain
<f.. 'Tli..,> =(9R+9w-+~B)A/..J3 (7.13a)
<l3'T1 i. 1 > =(~w-9:B)CIf2. (7.l3b)
(7.l3c)
The corresponding transition probabilities are given by

104 H. LIPKIN
where each expression is split into the direct terms considered in

the naive parton model and the interference terms normally neglected.
Note that when the charges are independent of colour, the expres-
sions (7.l3b), (7.l3c), (7.l4b) and (7.l4c) vanish and colour oc-
tet states cannot be excited from a colour singlet state, as is ex-
pected.
Equation (7.l4a) shows that when only colour singlet final

states are excited, the interference terms for all final states
have the same sign and cannot be neglected without causing a seri-
ous error. Thus, the naive parton model which neglects interfer-
ence terms cannot be used when only colour singlet states are exci-
ted.
Combining Eqs. (7.l4b) and (7.l4c) gives the total transition

probability for colour octet transitions
l<f31 Tli. .. >I1. + I<F,ITI ,.,>11.= 2'[(9:+ 9~+9!)

(7.14d)
-(9R9w+9w9s + 989ft) JC1 /3
The interference terms for transitions to colour octet final states
are seen to also have the same sign and give a non-negligible co-
herent contribution. However, the phase of these interference
terms is opposite to that of the transitions for the colour singlet
final states. Thus the condition for the naive parton model to hold
is that the interference terms from transitions to colour octet fi-
nal states must exactly cancel those from colour singlet final
states. From Eq. (7.14) this condition can be expressed
(7.15)
where the summation on the left-hand side is over all transitions

to colour singlet final states and on the right-hand side over all
transitions to colour octet final states.
Equation (7.15) shows that the naive parton model which ne-
glects coherence between the three diagrams of Fig. 7.1 is valid
only when there is a definite relation between the total cross-
sections for producing colour singlet and colour octet final states.
The exact value of the ratio of octet to singlet production depends
upon the values of the coupling constants, but ~s always of order
unity. For the Han-Nambu model this ratio can be seen to be ex-
actly unity by noting that the expressions (7.l4a) and (7.l4d) be-
come equal when the Han-Nambu coupling constants and the condition
(7.15) are used. Thus if the Han-Nambu model is correct, the naive
parton model predictions become valid only when the total cross-
sections for colour singlet and colour octet production become equal.
We now consider the possibility of observing a colour octet com-

ponent in the electromagnetic current in a second-order electromag-
netic deep inelastic process whose matrix element depends quadrati-
cally on the current. However, as we have seen in the meson decay
example, the validity of any second-order treatment depends upon a
model dependent factor often overlooked. If the underlying model
for the second order transition involves successive emission and/or
absorption of photons, the intermediate state between the two elec-
tromagnetic transitions must have an appreciable colour octet com-
ponent if the effects of the colour octet component of the current
are to be detected. Dynamical suppression factors for this colour
octet component may prevent the observation of such effects, as
shown above in the effect of intermediate state propagators in the
meson decays discussed in Section 7.2.
As an example of such a suppression, we note that the total

cross-section for photon absorption considered above is related by
the optical theorem to the imaginary part of the forward Compton
106 H. LIPKIN
amplitude. A calculation of this second-order amplitude without

consideration of the above arguments would include quadratic con-
tributions from the colour octet component of the current, which
had a colour singlet component and could give a non-vanishing ma-
trix element for the elastic scattering process. However, the dis-
CUSS10n of total absorption cross-sections shows that the contribu-
tion of such colour octet contributions to the absorptive part of
the amplitude must vanish as long as the energy is below the colour
threshold. A translation of equations (7.11) to (7.15) into the
language of forward Compton scattering shows that the propagator of
the intermediate state in the scattering process must be considered
very carefully and this propagator violates the conditions of the
naive parton model.
The essential features of the properties of the propagator can

be seen by noting that the optical theorem represents the absorp-
tion processes shown in Fig. 7.1 by squaring the amplitude. This
square includes not only the diagonal terms, like that shown 1n
Fig. 7.3a, in which Figs. 7.la, 7.lb and 7.lc are individually
squared but also the off diagonal terms, like the one shown in
Fig. 7.3b, in which the diagram of Fig. 7.la is joined to the con-
Fig. 7.3 Compton scattering 1n

quark parton model
jugate of the diagram of Fig. 7.lb or Fig. 7.lc. When this is ex-
pressed as a diagram for elastic Compton scattering it shows an in-
termediate state undergoing a colour change. Even though the con-
ditions of the naive quark model are assumed to hold and the same
quark which has absorbed the intial photon emits the final photon,
the colour of this quark can change during the intermediate state
as a result of the colour oscillations mentioned above. These os-
cillations can be studied in detail by examining the properties of
the propagator.
We denote the three states produced by the diagrams of Fig.

7.1 as [R>, [W> and [B>, respectively, corresponding to transitions
in which a red, white and blue quark absorbs the photon. These
states are linear combinations of the colour eigenstates [f 1 >, [f3>
and [fs>,
(7.l6a)
(7.l6b)
(7.l6c)
The colour eigenstates [f 1 > and [fs> have different energies El and
Es because of the energy required for excitation of colour octet
states. Thus if the state [R> is created at a time t = 0, the rela-
tive phase of the components [f 1 > and [fs> change with time and in-
troduce admixtures of the other states. For example,
(7.l7a)
(7.l7b)
The colour of a quark in the intermediate state thus takes

place at a frequency (Es - E 1 )/2.
108 H. LIPKIN
Thus the colour excitation threshold defines a time or energy

scale which determines whether a given process measures the charge
of a Han-Nambu quark or the average charge over the colour degree
of freedom. The colour is seen to change in the intermediate state
at a rate determined by Eq. (7.17). If the transition takes place
in a time short compared to this charge fluctuation time, then the
naive parton model result should be valid and give the charge of
the quark. If, however, the intermediate state lives a long time
compared to this fluctuation time, the charge is averaged over
colour, and the results are the same as that given by the fraction-
ally charged model.
The lifetime of the intermediate state is short, if it ~s do-

minated by high momenta; ~.e. by states which are high above the
colour threshold. Thus we see again that the relevant parameter is
the ratio of the energy of a typical intermediate or final state to
the threshold for colour excitations.
8. MIXING AND PSEUDOSCALAR MESONS
8.1 Introduction
Why is SU(3) such a good symmetry in some places and so badly

broken in others? Why are some hadrons good SU(3) eigenstates and
others badly mixed? The quark model seems to give part of the an-
swer. Mesons are quark-antiquark states and baryons are three-quark
states. This, plus isospin and hypercharge conservation, automati-
cally force most of the hadron states to be good SU(3) eigenstates.
Consider the n+, for example. This is the IS (ud) configuration.
There is no other (qq) state available with which it can mix without
violating isospin or hypercharge conservation or introducing larger
numbers of quarks. The n+ is thus a pure SU(3) octet state, even if
there is a large SU(3) violation in quark-quark interactions. The
same is true for all states in the lowest meson and baryon multi-
- - + +
plets (0 , 1 ,1 and 3/2 ) except for the I = Y = 0 mesons, where
we find n - n' mixing and w - ~ mixing. The general conclusion is

that SU(3) symmetry breaking is strong enough to mix any states
which are allowed to mix. But the quark model, which restricts
hadrons to qq and 3q, and isospin and hypercharge conservation
leave very few states which can mix.
Mixing can be described by perturbation theory in most cases.

If ~o is the unperturbed SU(3) eigenstate and ~. denotes the
~
states with which it can mix, the physical eigenstate in broken

SU(3) is
where V is the interaction responsible for the SU(3) symmetry

breaking and E. and Eo are the energies of the unperturbed states.
~
If the energy denominator is very small compared with V, then

degenerate perturbation theory must be used, and the interaction V
is diagonalized in the subspace of nearly degenerate states. This
occurs in the standard treatment of w - ~ mixing, for example.
Two kind of symmetry breaking terms are generally considered:
1) Mass terms. A flavour-dependent mass term for quarks seems

to be the dominant symmetry breaking mechanism for the vector
and tensor mesons. Diagonalizing the mass term gives a good
approximation to the physical eigenstates.
2) Loops. Hadron states can be mixed by transitions via inter-

mediate two-particle or multiparticle states. The loop dia-
grams describing these transitions are of two types, depending
upon the nature of the intermediate state.
a) Gluon loops. Since gluons are assumed to be flavour singlets,
all gluon intermediate states are flavour singlets, and they
110 H. LIPKIN
are connected only to flavour singlet hadron states by the

conventional gluon emission and absorption interactions.
Thus gluon loops do not break SU(3) in this approximation.
b) Hadron loops. Even if the three-point functions for coupling

a hadron to a two-hadron intermediate state is assumed to be
SU(3) invariant, these loop diagrams break SU(3) when the
physical masses are introduced for the propagators of the in-
termediate states.
8.2 The axial vector (Q) mesons
As an example of mixing by loops, let us consider the strange

axial vector mesons 1+ classified as 3 PI and Ipi in the quark model.
We denote the strange members of the Al and B octets by QA

and QB' respectively. The dominant decay modes K*TI and pK are
allowed for both QA and QB states. In the limit of SU(3) symmetry,
conserved "parities" G and G analogous to G parity can be defined
u v
by replacing isospin by U spin or V spin 1n the definition of G
parity. The neutral and charged Q's are eigenstates of Gu and Gv '
respectively. However, just as the K mesons are not eigenstates
. . .. .,'(
of G parity. Thus there 1S no select10n rule forb1dd1ng K TI and
pK final states for either of these decays. If the QA and QB are
produced coherently in some experiment, they contribute coherently
to the pK and K*TI final states 70 ).
If SU(3) is broken, G and G parities are not conserved.

u v
There can then be mixing, analogous to mixing, between the QA
w~
and QB states, even though G parity remains conserved and prevents

mixing of the corresponding non-strange states. However, there is
no ideal mixing angle determined by quark masses, as in the w
case, because the QA and QB have the same quark constituents and
are not mixed by a mass term. Some other SU(3) breaking mechanism
is needed to produce the observed mixing.
Consider the decay of the mixed states
(8.la)
(8.lb)
where e is the mixing angle.
For the K*n and pK decay modes the branching ratio is unity
in the SU(3) limit except for differences in kinematic (phase
space) factors for the two final states. However, because the
two octets have opposite charge conjugation behaviour, the Al oc-
tet decay is described with F coupling and the B octet decay with
D coupling. The relative phases of the Kp and K*n decay amplitudes
are thus opposite for the two cases
(8.2a)
(8.2b)
the decay amplitudes for the mixed states (8.1) are then
<K*1C I Q .. '> = (.OS 8- < K-1t I (;lAo> + Si.flS< Klflt I QB> (8.3a)
<Ke IG .. ) :_t.o~9<KlIt1t1 ~A>+Si... f).<I<1tIGB) (8.3b)
<K*1t IG)2> =_Si",9-< K*1t I GA>+ (.059-( K*1t ( QB> (8.3c)
<Ke IG2 >= Si.-nf}(I<*1t1 G,,>+c.os&<l<lIF1tIG B> (8.3d)
Equations (8.3) show that for any mixing with a real phase,
the effect for one eigenstate is to enhance the K*n decay mode and
112 H. LIPKIN
and suppress the K , and vice versa for the orthogonal eigenstate.
For = 45 0 , we obtain
I(KeIG.f>I2. = I<KTC1Gz.>I2.
1(l(iJ1t IG .. >I2. If2. - (8.4a)
= I(K lJ 1tIG A>-(I<-1tIGa>I L

I<K-Jtl GA> +( K-1tIG a >I'
Thus QJ is decoupled from Kp and Q2 .
~s decoupled from K*n. The de-
coupling is exact for the fase where the QA and QB states are
equally coupled to the K*n mode and is still a good approximation
over a side range of couplings. For example, as long as
of >
I... I ( K-lrl GA 12.
(8.4b)
4 - I ( K-1t 1Q S >I ~
we s ti 11 have
,< Kf , G.. >12 _1< 1(-1t 1c:i1.>1lo L .. (8.4c)

I<K1'tIGi~12 -'(I I GZ >'2. 9
A dynamical mechanism which naturally leads to this mixing is

the SU(3) breaking in decay channels originally introduced to ex-
plain 71 ) w~ mixing before SU(6) and the quark model. The states
QA and QB are coupled to one another via their decay channels K*n
and Kp:
(8.Sa)
(8.Sb)
In the SU(3) symmetry limit, the two transitions (8.5a) and

(8.5b) exactly cancel one another and produce no mixing. This
cancellation no longer occurs when SU(3) breaking introduces kine-
matic factors arising from the mass difference between the two
intermediate states. These suppress the strength of the transi-
tion (8.5b) via the higher mass Kp intermediate state relative to
the transition (8.5a) via K*TI.
The simple analysis of the trans~t~ons (8.5a) and (8.5b) gives

45 o mixing for the eigenstates if <K*TIIQA> = <K *TI1QB>. This de-
couples the two states from K*TI and Kp, .
respect~vely. However, a
more careful analysis shows that two partial waves are present ~n
the decay, s wave and d wave, and the result is very sensitive to
the relative amplitudes and phases of the sand d waves. In par-
ticular, for the ratio of s to d wave amplitudes predicted by the
naive SU(6)w quark model, the transitions (8.5) vanish and cannot
produce mixing, because the QA is coupled only to vector meson
states with transverse polarization and the QB is coupled only to
longitudinally polarized states 70 ) . For this reason, the mechanism
(8.5) for mixing was dropped.
A recent analysis of the experimental properties of the Q

mesons suggests a mixing of SU(3) eigenstates with a 45 0 mixing
angle with one of the eigenstates decaying only to K*TI and not to
Kp and vice versa for the other state 72 )
Now that the SU(6)W predictions are known not to agree with
.
exper~ment
73) ,part~cularly
. . I I I
~n the c ose y re ate
d po I ar~zat~on
. .
predictions for B and Al decays, and the experimental data are con-
sistent with pure s wave for the Q decays, the mixing mechanism
(8.5) should, perhaps, again be considered. However, a more real-
istic calculation would consider the coupled channels K*TI and Kp
through the resonance region, with phase space factors changing
within the resonances because of the proximity to threshold.
114 H. LIPKIN
8.3 Troubles with pseudoscalar mesons
The vector and tensor meson nonets are well-described by at-

tributing all the SU(3) symmetry breaking to a flavour-dependent
quark mass term, and assuming nonet degeneracy except for this
mass term. Many experimental predictions of this description have
been successfully tested. However, the analogous predictions do
not work for the pseudoscalar mesons. At first, it was assumed
that some additional interaction could change the mixing angle
from the so-called ideal mixing produced by the quark mass term,
and phenomenological predictions were made in which the mixing
angle was left as a parameter to be determined from experimental
data. However, these are also in disagreement with experiment.
The conventional mixing description seems to be in both ex-

perimental and theoretical trouble for the pseudoscalar mesons.
The nand n' do not behave like orthogonal mixtures of a single
SU(3) singlet and a single SU(3) octet. More complicated mixing
~s indicated perhaps requiring inclusion of radially excited states
. . 23,74)
as we 11 as ground-state conf~gurat~ons .
The use of the quark model to determine the mixing angles of

neutral mesons from experimental data on neutral meson production
processes was first suggested by G. Alexander 75 ) . This work,
52 76)
based on the Leven-Frankfurt additive quark model' in which
every hadron transition is assumed to involve only one active quark
with all remaining quarks behaving as spectators, presented a num-
ber of predictions which have since been shown to be in very good
agreement with experiment. These include the first derivation of
the A... Z rule for four-point functions, as the prediction that ~
production is forbidden in TIN reactions since the process requires

two active quarks in the same hadron. Also obtained were the pre-
diction of no exotic t channel exchanges and some sum rules and
equalities which are listed below. Analysis of a decade of experi-
mental data shows a consistent pattern of good agreement with all
predicted relations for processes of vector meson production and

strong disagreement with relations for processes of pseudoscalar
meson production, particularly for relations involving n' produc-
tion. We suggest that an appropriate conclusion from these re-
sults is that the quark model description indeed holds for these
processes, but that something is wrong with the pseudoscalars,
particularly the n'.
The relevant sum rules are the charge exchange sum rule (CHEX)
a- ( It - P ~ n: 11. ) + G' (n;- p~ "l,"'-) + G'(n.- p ~ "11 'n) =

(8.6a)
:: a-( K+It'l-+KOp)+a-(I<-p .... ROI)'l)
and the strangeness exchange sum rule (SEX)
CT ( K- P .... 'rf Y) + a- ( I( - p ... ")' Y) =

.. <r(I<-p~TCOY)+Q"(Tt-p 4- KOY) (8.6b)
These sum rules hold for any meson nonet and do not make any
assumption about the mixing angle, except for the conventional
description of the nand n' as two orthogonal linear combinations
of pure SU(3) singlet and octet states defined in terms of a single
mixing angle. For the case of ideal mixing, as in the vector me-
sons, the two sum rules each split into two equaltities, CHEX be-
comes:
(8.7a)
which is just the A.. Z rule, and substituting (8.7a) into (8.6a)
gives
CTln-.p ~ eO~)+CS"(Tt-p~W't1.) =
: er( K+"., + J(1tp)+CS"'CK-p ~k'''0IP1.) (8.7b)
116 H. LIPKIN
With ideal mixing SEX becomes
(8.8a)
(8.8b)
The relation (8.8a) is seen also to be a consequence of the A Z

rule for the meson vertex. The incident K contains no d for d
quarks or antiquarks and therefore produced via the uu component
which is a linear combination of the two with equal weight.
If there is no mixing, which is a rough approximation for the

pseudoscalar mesons, the charge exchange sum rule simplifies to
a (u-p .. It 'n.) + 3 crt yt- {) ... "Ia'rl) =

(8.9a)
= 0'"( I<+~ ..... 1<0,) + cs-( K-p -+ ifo'rl)
(8.9b)
All the vector meson relations (8.7) and (8.8) are in excel-
lent agreement with experiment. However, the pseudoscalar meson
relations (8.6) and (8.7b) are in strong disagreement. The rela-
tion (8.7a) agrees with experiment if the n is assumed to be pure
octet. This suggests that the conventional picture in which there
is small mixing may be valid for the n, but that something is
wrong with the nt, and it is wrong in the direction that the n'
has an inert piece in the wave function which does not contribute
to the sum rules (8.6) and (8.7).
More recent evidence of trouble in the n - n' system comes

from data presented at this conference on neutral meson production
in K-p reactions at 4.2 GeV/c. The previously observed trouble
with the SEX sum rule (8.lb) is confirmed with higher statistics.
In addition, there may be difficulty with backward production.

Okubo 77 ) has pointed out that conventional mixing predicts that
the ratio of n' to n production must be a universal constant in
all processes where there are no active strange quarks,
0'"( A + B ~ ~ I + X )
(8.10)
cr(A+B~1tf + X)
where A, Band S do not contain strange quarks and denotes the

ns
particular linear combination of nand n' which contains only non-
strange quarks, the analogue of the physical w. Okubo finds a
value of K = 0.5 0.25 by analysis of a large number of processes.
.
But recent experiments 7 8 )give
.
(8.11)
for the ratio of n' to n production in the backward direction in

Kp + An or An'. If this is a baryon exchange process, the coup-
ling of the nand n' to non-strange baryons should also go via the
nns component and the processes should satisfy Okubo's universality
relation (8.10). The fact that K > 1 for this case whereas K < 1
for all meson exchange processes investigated by Okubo might sug-
gest a difference between meson exchange and baryon exchange, if
the discrepancy of less than two standard deviations proves to be
statistically significant. This is again consistent with the
description of the n as having a piece in the wave function which
does not contribute to the meson exchange sum rules because of
poor overlap with the wave function of the incident meson. For
baryon exchange there is no such overlap integral and the addition-
al piece could have a sizeable contribution, giving a higher n'
production cross-section relative to the n than in meson exchange
reactions.
118 H. LIPKIN
We suggest that there is indeed an additional piece in the

n' wave function and that it is a radially excited configuration.
This leads to a re-examination of the standard mixing folklore
and the .
d~scovery .
that ~t .
~s complete I ' 'f'~e d 74 ) .
y unJust~ In a
formulation which begins with unperturbed singlet and octet states
in the SU(3) symmetry limit, there is no reason to assume that
SU(3) symmetry breaking should admix only the lowest ground states
of the singlet and octet spectra. This may work for the tensor
and vector mesons, where the entire nonet seems to be degenerate
in the SU(3) symmetry limit and the dominant breaking of nonet
symmetry is by a quark mass term. The degeneracy suggests the use
of degenerate perturbation theory which diagonalizes the symmetry
breaking interaction in the space of the degenerate unperturbed
states. The mass term has no radial dependence and would not mix
ground state and radially excited wave functions which are ortho-
gonal and would have a zero overlap integral.
For the pseudoscalars where there is a large singlet-octet

splitting in the SU(3) symmetry limit there is no reason to use
degenerate perturbation theory and mix only ground state wave func-
tions. Furthermore, the singlet-octet splitting can only be pro-
duced by an interaction which violates the A... Z rule because it
is not diagonal in the quark basis and m~xes ss with u~ and dd.
The accepted mechanism for such A.. Z violation in the pseudosca-
lars is annihilation of the quark-antiquark pair into gluons and
the creation of another pair. Here there is no reason to restrict
the pair creation to the ground state configuration. There is no
overlap integral between the two qq states, as the intermediate
gluon state does not remember which radial configuration it came
from. If the annihilation process depends primarily on the value
of the qq wave function at the origin, then all radially excited
configurations couple with equal strength for wave functions from
a confining linear potential.
Thus there is considerable reason to suspect that the trouble

with pseudoscalar meson sum rules is in admixture of a radially
excited wave function into the n'. One might expect the n to be
purer because the SU(3) flavour octet state does not couple to
gluons which are singlets and because it is the lowest state, far
in mass from the nearest SU(3) singlet radial excitation. The n',
on the other hand, is sitting in between the ground state and first
radially excited octet states and would be expected to mix with
both. Note that mixing of the octet ground state and first radial-
ly excited octet state by an SU(3)-symmetric potential need not
be considered because it is merely a change in the radial wave func-
tion. This mixing can be transformed away by choosing a new radial
basis (i.e., a slightly different potential) for which the modified
ground wave function in the original basis is the exact ground
state in the new basis.
9. WHY ARE THERE MYSTERIOUS REGULARITIES

IN HADRON TOTAL CROSS-SECTIONS?
9.1 Flavour dependence of hadron total cross-sections
The very ,
prec~se ,
exper~mental d ata 79) now ava~'I a bl e on '
p~on,
kaon and nucleon total cross-sections give us some information

about the difference between the interactions of strange and non-
strange particles with matter. Careful examination of the data
show very clearly that there is a difference between strange and
non-strange particles and that there are puzzles not explained by
the quark model. This is strikingly shown in linear combinations
of cross-sections which have no Regge component and are therefore
+
conventionally assumed to be pure Pomeron. The K p and pp chan-
nels are exotic and have no contribution from the leading Regge
exchanges under the common assumption of exchange degeneracy. The
following linear combinations of meson-nucleon cross-sections are
constructed to cancel the contributions of the leading Regge tra-
jectories:
120 H. LIPKIN
(9.la)
(9.lb)
Figure 9.1 shows these two quantities on the conventional plot of

cross-section versus Plab on a log scale.
201~----~----------------~r-------~------~
15 f(j(~P)
~~~~---
~
~
.c
(pK)
E
10
b
O~ __~~____~~______~~____~~____~
6 200
Plab (GeV/c)
Fig. 9.1 Plots of Eqs. (9.1) and (9.2)
o( ) as defined by Eq. (9.la) is the quark model expression

p
for o(); i.e., the cross-section for the scattering of a strange
p
quark-antiquark pair on a proton. The very simple energy behaviour
of this quantity as seen in Fig. 9.1 is striking. It shows a mono-
tonic rise beginning already at 2 GeV/c.
The quantity 6(rrK) , defined by Eq. (9.lb) represents the dif-

ference in the scattering of a strangeparticle and a non-strange
particle on a proton target. In the quark model this is the dif-

ference between the scattering of a strange quark and a non-strange
quark on a proton target after the leading Regge contributions have
been removed. This difference between strange and non-strange
also has a very simple energy behaviour, decreasing constantly and
very slowly (less than a factor of 2 over a range P lab of two
orders of magnitude). So far there is no good explanation for why
strange and non-strange mesons behave differently in just this way.
Since the two quantities (9.1) have no contribution from the

leading Regge trajectories they represent something loosely called
the Pomeron. However, their energy behaviours are different from
one another and also from that of the quantities a(K+p) and a(pp)
which should also be "pure Pomeron". However, the following linear
+
combinations of a(K p) and a(pp) have exactly the same energy be-
haviour as the meson-baryon linear combinations (9.1)
(9.2a)
(9.2b)
These quantities are also plotted in Fig. 9.1.
The equality of the quantities (9.2) and the corresponding

quantities (9.1) suggest that the Pomeron, defined as what is left
in the total cross-sections after the leading Regge contributions
are removed by the standard prescription, consists of two compo-
nents, one rising slowly with energy and the the other decreasing
slowly. The coefficients in Eq. (9.2) were not picked arbitrarily
but were chosen by a particular model. In this model the rising
component of the total cross-section is assumed to satisfy the
standard quark model recipe exactly:
122 H. LIPKIN
(9.3a)
where Y denotes a A or L hyperon. The falling component has been

assumed to satisfy the following relation
()2. (Kp) :: ! <S"2(ltP): ~a'"'2(PP)::
2. 9
:' .! CS'z. (Yp) =l:. 0""2 (8 p)
(9. 3b)
a 3
This particular behaviour is suggested by a model in which the cor-
rection to a simple quark-counting recipe comes from a double ex-
change diagram involving a Pomeron and an f coupled to the incident
. eo)
particle .
We thus see unresolved problems in the total cross-section

data associated with the questions of what is the difference be-
tween strange and non-strange particles and what is the nature of
the Pomeron. Note that Eq. (9.lb) defines the difference between
the scattering of a non-strange quark and a strange quark while
Eq. (9.2b) can be interpreted as the difference between the scat-
tering of a quark in a baryon and a quark in a meson. The fact
that the strange-non-strange difference and the meson-baryon dif-
ference are equal and have the same energy behaviour over such a
wide range is a puzzle which may be explained by Pomeron-f double
exchange but may also indicate something deeper.
9.2 The two-component Pomeron formula
A very good fit to the experimental total cross-section data

up to 200 GeV/c has been obtained with the two components (9.2)
and (9.3) parametrized by simple power behaviour. This gives a
80)
formula with five parameters which were adjusted to fit the data
(9.4)
where C1 = 6.5 mb, C2 = 2.2 mb, and CR = 1.7 mb,
(9.5a)
a-2 (Hp) = N~ N~s (P~/20)- s (9.5b)
.,
CTR (Hp) = (N~ +2.N;)( P&llof I (9.5c)
NH is the total number of quarks and antiquarks in hadron H (N H 2

q H q
for mesons and 3 for baryons), N is the total number of non-
ns H H
strange quarks and antiquarks in hadron Hand N= and N= are the
_ _ n p
total number of nand p antiquarks in hadron H, E = 0.13 and 0 =
0.2.
The dependence of the individual terms in Eqs. (9.5a) and

(9.5b) on the quantum numbers of H are determined by the model and
discussed ~n Ref. 80. The explicit form for the energy dependence
is chosen to m~n~m~ze the number of free paramters. Thus, power
behaviour is chosen rather than logarithmic for the two components
of the Pomeron, because two param~ters are sufficient to describe
a power and at least three are needed to describe logarithmic be-
haviour. The Regge term was chosen to minimize the number of
free parameters by assuming exact duality and exchange degeneracy
for the leading trajectories with the conventional intercept of
one-half.
The formula (9.4) predicts that plots of Gtot (Hp) (P lab /20)
o
(E+O)
vs Plab should show straight lines for all cross-sections and
linear combinations of cross-sections which have no Regge contri-
bution. This is strikingly verified in Figs. 9.2a and 9.2b, which
show straight lines for Gt (K+p), 0' (pp), and for the linear
ot tot
combinations (9.1) and (9.2). A straight line is not obtained for
Gtot(n-p), which has a Regge component.
124 H. LIPKIN
However, when this Regge contribution is removed by

plotting 0 tot (n-p) - 0 R (n-p), as defined by Eq. (9.5c) another
straight line appears.
The formula (9.4) predicts that the straight lines for

- - +
(2/3)0 tot (pp), 0 tot (n p) - 0 R (n p), 0 tot (K p), and 0(~p) = 0l(pK)
should have the same slope and be equally spaced. This is clearly
shown also in Fig. 9.2a. The straight lines for 6(nK) and 6(MB)
are predicted to have zero slope and a value equal to the spacing
between the equally spaced parallel lines. This is in qualitative
agreement with Fig. 9.2a), although there is a slight rise, sug-
gesting that the value of 0.2 for the parameter 6 is a bit too
high. Similar straight lines are obtained with slight variations
of the parameters. Changing 6 to 0.185 gives a better fit to the
data.
The extension of the formula (9.4) to the real part of the

amplitude is a straightforward application of analyticity and
crossing, which is particularly simple for terms with power be-
haviour 8l ) and gives the following expression for the ratio of the
real to the imaginary parts of the Hp amplitude
(9.6)
9.3 Fits to higher energy of the two-component

Pomeron formula
The total proton-proton cross-section and the real part of the

82)
forward scattering amplitude have been recently measured at the
ISR. Table 9.1 shows that the new data in the energy range equiva-
lent to Plab = 500 to 2000 GeV/c are in excellent agreement with
predictions from the five-parameter formula (9.4)-(9.6) with no

N
0....... ~O b.
u b. b. b.
.......
0 b.
> Q)
b.

..
t.:>

.0
30 0
b.b.
oo~ b.

~
II
Ob.b.
E ~ /:,b. i
N II !i!
allo ,...
....
0_ b.b.
II
o 20 0 ,,_ill !1
,,-
N
_II-
.............

.0
0 "

(1.-
ti llV .....
-
~
10 II
" ~~
0
b- !j!!i! ljl ~ljl_ II ~II-"e .. It III tI~ ti ti
0
0 2 3 4 5 6
0.33 0.33
(~ab ) (G eV Ie)
Fig. 9.2a 0tot(Hp) x (Plab/20)O.2 plotted against (P lab )O.33:
/:, tot(n-p)
0 (2/3)Otot(pp)
tot(n-p) - R(n-p)
+
'- tot(K p)
x (<jJp)
1/ ol(pK)
0 /:'(HB)
+ /:, (nK)
t..)
0.
Table 9.1
Theoretical predictions and experimental data for tot (pp) and p(pp)
P1ab IS 0 tot (PP) 0 tot (Pp) p(pp)
Theory Theory Experiment

Theory Experiment
(GeV/c) (GeV) (mb) (mb) (mb)
498 30.6 41.8 40.0 40.1 0.4 0.025 0.042 0.011

,
1064 44.7 42.8 41.6 41.7 0.4 0.064 0.062 0.011 I
I
1491 52.9 43.5 42.5 42.4 0.4 0.079 0.078 0.010 I
i
2075 62.4 44.3 43.5 43.1 0.4 0.092 0.095 0.011
4600 92.9 46.8 46.2 47.0 0.8 0.118
10000 137.0 49.8 49.5 50.6 1. 2 0.138
25000 217.0 54.3 54.0 53.8 2.2 0.156
40000 274.0 56.9 56.7 55.0 3.0 0.163
100000 433.0 62.7 62.6 J:

--- -- - _ . _ - _ . - ~--~ '------- r
-g
A
Z
adjustment of the values of these parameters from already published

values fixed by fits to data below 200 GeV/c. Table 9.1 also lists
predictions for higher energies and shows remarkable agreement with
results from cosm~c
ray exper~ments
83) up to Plab = 4 0,00 GeV c. /
The plots of Fig. 9.2a are extended to these higher energies in
Fig. 9.2b and show a good straight line with the same parameters.
An equally good straight line is obtained if 0 is changed to
0.185. Whether these agreements confirm the validity of the over-
simplified two-component model is unclear. However, the formula
can certainly be used as a simple parametrization of the data and
a guide to the physics of further experiments. The ISR group fit
their data with a seven-parameter formula 82 )
300 I I I
'"
0_
()
.......
>.,
250 f -
(!)
I
200 - -
.a
E I
'"0_ 1501- -
0
C\I

..............
.a 1001- -
c
0.-

'0 50 1-/ / -
b-
I I I
Oo 10 20 30
0.33 0.33
(Plab ) (GeV Ie)
(P )0.33
Fig. 9.2b (Jtot(pp) x (P lab /20)2 plotted against lab
The good fits obtained to very-high-energy data indicate that

these rather crude approximations are nevertheless adequate up to
these energies. As long as this reasonable fit continues, models
containing more detailed assumptions will not be easily tested by
128 H. LIPKIN
the available data. For example, as long as a good fit ~s ob-

tained with power behaviour for the first component the necessity
for logarithmic terms will be difficult to demonstrate since a
considerably better fit is required to justify the use of addition-
al parameters. The same is true for more detailed or realistic
descriptions of the Regge component, since breaking exchange de-
generacy or choosing a value different from one-half for the inter-
cept necessarily requires more parameters. However, as soon as
data apper which fail to fit this formula, the underlying assump-
tions are so simple that the physics of the disagreement should be
readily apparent. The nature of the disagreement might suggest,
for example, that the rise of the cross-sections is logarithmic
rather than a power, that exchange degeneracy is breaking down, or
that the Regge intercept is not one-half. There may also be a
breakdown of the two-component Pomeron picture if the dependence
on the quantum numbers of hadrons H no longer satisfies the simple
relations of the model. Thus, regardless of the validity of the
two-component Pomeron description, the formula (9.4) should be a
valuable guide to the analysis of data on high-energy total cross-
sections and real parts of scattering amplitudes.
9.4 Lipkin's Crazy Parton Model
Another puzzle is suggested by the fit to the data with 0 =
0.185 and E = 0.13, which satisfy the condition 0 = (0.5 - E)/2.

This condition suggests that the total cross-section is the square
of an amplitude with two components, one varying as P l / b2 and one
-1/4 a
varying as P lab . Then 01 and OR represent the squares of these
components and 02 is the interference term. It is tempting to try
to fit this regularity with a parton model in which the total cross-
section is assumed to come from two contributions: 1) a "Pomeron
diagram" in which a quark in the beam exchanges a "Pomeron" with
the proton target, and then fragments into the final state; 2) a
"Reggeon diagram", in which a Reggeon is exchanged before fragmen-
tation into a different final state. The new decreasing component
O2 (Hp) might arise from interference between Pomeron and Reggeon

amplitudes. One might even explain this interference by invoking
"f-dominance of the Pomeron" to show that the same complicated
final states produced by fragmentation after Pomeron exchange can
also be produced after f exchange, and therefore the two must be
coherent and interfere.
The dependence of the Pomeron and Regge diagrams on the quan-

tum numbers of the beam particle are exactly those required by
Eqs. (9.Sa) and (9.Sc), while the interference term naturally has
the flavour dependence of the Pomeron-f double exchange of Eq.
(9.3b) which leads to Eq. (9.Sb).
STUDENTS: Something is obviously wrong with this model.

See if you can guess it before reading further!
Unfortunately, the "interference term" 02(Hp) also exists in

channels like pp and K+p which have no Regge term. Thus the ~n
terference model is in contradiction with elementary quantum
mechanics. We are left with the puzzle:
Why can hadron total cross-sections be fit with three compo-

nents having the energy dependence of a slowly rising Pomeron,
a decreasing Regge exchange and Pomeron-Regge iterference, when a
non-vanishing interference term is present in some cases where the
direct Regge term vanishes?
130 H. LIPKIN
REFERENCES
1) F. Gursey, Erice Lectures 1977.
2) H.J. Lipkin, in Symmetry Principles at High Energy, Fifth

Coral Gables Conf., eds. A. Perrnutter, C.A. Hurst and
B. Kursunoglu (Benjamin, 1968), p. 261.
3) S. Weinberg, Phys. Rev. Letters 19 (1967) 1264.

A. Salam, in Elementary Particle Theory, ed. N. Svartholm
(Almquist and Forlag, Stockholm, 1968).
E.S. Abers and B.W. Lee, Phys. Reports 9C (1973) 1.
4) B.J. Bjorken and S.L. Glashow, Phys. Letters 11 (1964) 225.
5) S.L. Glashow, J. Iliopoulos and L. Maiani, Phys. Rev. D2 (1970)

1285.
6) D.C. Cundy, in Proc. XVII Int. Conf. on High Energy Physics,

London (1974), ed. J.R. Smith, p. IV-13l (1974) 190.
7) G. 't Hooft, Nuclear Phys. B33 (1971 173; B35 (1971) 167.
8) o.w. Greenberg and C.A. Nelson, Phys.Reports 3C (1977) 69.
9) S. Coleman, Erice Lectures (1977). These proceedings.
10) G. 't Hooft, 1972 Marseilles Conf. on Yang-Mills fields

(unpublished).
D.J. Gross and F. Wilczek, Phys. Rev. Letters 30 (1973) 1343.
H.D. Politzer, Phys. Rev. Letters 30 (1973) 1346 and Phys.
Reports l4C (1974) 129.
R.J. Crewther, CERN preprint TH.2ll9 (1975) to appear in
Proc. 1975 Cargese Summer Institute on Weak and Electromag.
Interactions.
11) H. Harari, to be published in the Proc. 1977 Experimental

Meson Spectroscopy Conf., Northeastern University (1977).
12) F. Iachello, Erice Seminar 1977. These proceedings. See

also P. Fayet and S. Ferrara, Phys. Reports 32C (1977) 249.
13) H.J. Lipkin, Fermilab-Conf-77/65-THY, to be published in Proc.

1977 Experimental Meson Spectroscopy Conf., Northeastern
University (1977).
14) R.H. Dalitz, Proc. Xlllth Conf. on High Energy Physics,

Berkeley (1966), Univ. of California Pre~s (1967) p. 215.
15) P. Federman, H.R. Rubinstein and I. Talmi, Phys. Letters 22

(1966) 208.
16) A. Wattenberg, to be published in Proc. 1977 Experimental Meson

Spectroscopy Conf., Northeastern University (1977).
17) H.J. Lipkin, Fermilab-Conf-77/66, to be published in Proc. of

Brookhaven Symposium on Prospects of Strong Interaction
Physics at Isabelle, April (1977).
18) A. Zichichi in High Energy Physics, Proc. of the 1975 EPS

Conference, Palermo, Italy, ed. A. Zichichi, Editrice
Compositori Bologna (1976) p.8.
19) H.J. Lipkin, Phys. Rev. Letters 28 (1972) 63 and Particle

Physics (Irvine Conference, 1961), ed. M. Bander
G.L. Shaw and D.Y. Wong (American Inst. Phys. N.Y. 1972)
p. 30.
20) J.J. Aubert et al., Phys. Rev. Letters 33 (1974) 1404.

J.O.E. Augustin et al., ibid. 33 (1974) 1406.
Comprehensive reviews have been given recently by G.J. Geldman,
in Proc. of the Summer Institute on Particle Physics, ed.
Martha C. Zipf (SLAC Report No. 198, 1976), p. 81.
F.J. Gilman, in Particles and Fields-'76, ed. H. Gordon and
R.F. Peierls (BNL, Upton, N.Y., 1977), p. B19.
R. Schwitters, ibid. p. 33.
G.H. Trilling, in Proc. of the Summer Institute on Particle
Physics, ed. Martha C. Zipf (SLAC Report No. 198, 1976)
p. 437.
B.H. Wiik and G. Wolf, DESY 77/01, Lectures given at the 1976
Les Houches Summer School.
21) Mary K. Gaillard, B.W. Lee and J.L. Rosner, Reviews of Modern
Physics, 47 (1975) 277.
22) S. Okubo, Phys. Letters 5 (1963) 165.

G. Zweig, unpublished (1964), and in Symmetries in Elementary
Particle Physics (Academic Press, N.Y. 1965), p. 192.
J. Iizuka, Supplement to Progress of Theoretical Physics,
37-38 (1977) 21.
23) H.J. Lipkin, in Deeper Pathways in High Energy Physics,

Proc. of the Orbis Scientiae 1977 Coral Gables Conf.
ed. B. Kursonoglu, A. Perlmutter and L.F. Scott, Plenum
Press, N.Y. (1977) p. 567.
24) The rule forbidding ~ + p~ and other stuff has been credited
to Okubo, Zweig, Iizuka and others in~rious combinations.
To avoid arguments about credit, one can refer to the A Z
132 H. LIPKIN
rule and allow the reader to insert the names of all de-
sired friends from Alexander to Zweig. For a general re-
view, see, H.J. Lipkin, in New Fields in Hadron Physics,
Proc. XI Rencontre de Moriond, ed. J. Tran Thanh Van,
CNRS (1976), p. 327 and Lectures at the Erice Summer School
(1976), Fermi1ab preprint Conf-76/98-THY to be published
in the Erice Proceedings.
25) H.J. Lipkin, Phys. Letters 60B (1976) 371.
26) B.W. Lee, in Proc. XVI Int. Conf. on High Energy Physics,
Chicago-Batavia, Ed. J.D. Jackson and A. Roberts (1972),
Vol. 4, p. 249.
27) H. Georgi and S.L. G1ashow, Phys. Rev. Letters 28 (1972) 1494,
Phys. Rev. D7 (1973) 561.
28) J.D. Bjorken in Proc. XVI Int. Conf. on High Energy Physics,
Chicago-Batavia, ed. J.D. Jackson and A. Roberts (1972)
Vol. 2, p. 299.
29) J. I1iopou10s in Proc. XVII Int. Conf. on High Energy Physics,

London (1974), Ed. J.R. Smith, P. 111-89.
30) D. Cohen et a1., Phys. Rev. Letters 38 (1977) 269.
31) C.W. Aker10f et a1., Phys. Rev. Letters 39 (1977) 861.
32) S. Herb et a1., Phys. Rev. Letters 39 (1977) 212.
33) E. Eichten and K. Gottfried, Phys. Letters 66B (1977) 286.
34) M. Jacob, "The double pomeron process" ISR Discussion meeting

between Theorists and Experimentalists, CERN preprint
June (1977).
35) C. Rosenzweig, Phys. Rev. D13 (1976) 3080.
36) W. Braunschweig et a1., Phys. Letters 67B (1977) 233.
37) c. Quigg and J.L. Rosner, Hadronic decays of the n , FERMILAB-

Pub-77/40-THY, to be published in Physics Lettefs.
38) H. Schopper, Erice Lectures 1977. These proceedings.
39) R. Brandelik et a1., DESY report 77/44.
40) D.W. Greenberg, Phys. Rev. Letters 13 (1964) 122.

WHYISTHERECHARM,STRANGENESS,ANDCOLOUR? 133
41) Y. Nambu, in Preludes in Theoretical Physics, eds. A. de

Shalit, H. Feshbach and L. Van Hove (North-Holland Publ.
Co., Amsterdam, 1966) p. 133.
42) S.L. Adler, Phys. Rev. 177 (1969) 2426.
43) J.E. Augustin et al., Phys. Rev. Letters 34 (1975) 764.
44) H.J. Lipkin, Quarks, Partons, Triality, Exotics and Coloured

Glue, in Proc. of Summer Inst. on Particle Physics,
SLAC, 9-28 July 1973, Vol. I (National Technical Informa-
tion Service, 1973), SLAC-167, pp. 239-253.
46) A. Chodos, R.L. Jaffe, K. Johnson, C.B. Thorn and V.F.

Weisskopf, Phys. Rev. D9 (1974) 3471.
W.A. Bardeen, M.S. Chanowitz, S.D. Drell, W.Weinstein and
T.M. Yan, Phys. Rev. Dll (1975) 1094.
47) J.C. Pati and C.H. Woo, Phys. Rev. D3 (1971) 2920.
R.L. Kingsley, Phys. Letters 40B (1972) 387.
See also Ref. 8, p. 94.
48) E.' Eichten et al., Phys. Rev. Letters 34 (1975) 369.
49) G. It Hooft, Nuclear Phys. B75 (1974) 461.
50) C. Quigg and J.L. Rosner, FER}lILAB-Pub-77/82-THY.
51) S. Okubo, Phys. Rev. D13 (1976) 1994 and D14 (1976) 108.
52) H.J. Lipkin, Phys. Reports 8C (1973) 173.
53) F. Gilman, M. Kugler and S. Meshkov, Phys. Rev. D9 (1974) 715;

Phys. Letters 45B (1973) 481.
54) J.L. Rosner, Phys. Rev. Letters 21 (1968) 950.

P.G.O. Freund, R. Waltz and J. Rosner, Nuclear Phys. B13 (1969)
237.
55) R.L. Jaffe, Phys. Rev. D15 (1977) 267 and 281.
56) S. Meshkov in Proc. 3rd Coral Gables Conf. on Symmetry

Principles at High Energy, eds. B. Kursonoglu,
A. Perlmutter and I. Sakmar (W.H. Freeman and Company,
San Francisco, 1966) p. 150.
57) R.H. Dalitz, in Proc. Oxford Int. Conf. on Elementary Particles

(Rutherford High Energy Lab., Chilton, January 1966) p. 157.
134 H. LIPKIN
58) A. De Rujula, H. Georgi and S.L. Glashow, Phys. Rev. D12

(1975) 147.
60) R.L. Jaffe, Phys. Rev. Letters 38 (1977) 195.
61) R.D. Carlitz, S.D. Ellis and R. Savit, Phys. Letters 68B (1977)
443. See also N. Isgur and G. Karl, Phys. Letters, in
press.
62) J. Rosen, private communication.
63) For a review, see A. Bodmer, in High Energy Physics and

Nuclear Structure, Proc. of the 1967 Rehovot Conf., ed.
G. Alexander, North-Holland Publishing Co., Amsterdam
(1967) p. 60.
64) K. Gottfried, to be published in the Proc. 1977 Hamburg Conf.
65) H.J. Lipkin, Phys. Rev. Letters 28 (1972) 63 and Particle

Physics (Irvine Conf. 1971), ed. M. Bander, G.L. Shaw
and D.Y. Wong (American Inst. of Phys., N.Y., 1972) p. 30.
66) H.J. Lipkin, in High Energy Physics, Proc. 1975 EPS Conf.
Palermo, ed. A. Zichichi, Editrice Compositori Bologna
(1976) p. 609.
D. Faiman, H.J. Lipkin and H.R. Rubinstein, Phys. Letters
59B (1975) 269.
67) M.Y. Han and Y. Nambu, Phys. Rev. 139 (1965) Bl006.
68) M. Chanowitz, in Color Symmetry and Quark Confinement, Proc.

XII Rencontre de Moriond, Ed. Tran Thanh Van, (1977), p. 25.
69) N. Colourbibbo, private communication.
70) H.J. Lipkin, Proc. Heidelberg Int. Conf. on Elementary Par-

ticles, Ed. H. Filthuth, North-Holland Publishing Co.,
(1968), p. 253 and Phys. Rev. 176 (1968) 1709 and FEID1ILAB-
Pub-77 / 84-THY .
71) A. Katz and H.J. Lipkin, Phys. Letters 7 (1963) 44.
72) G.W. Brandenburg et al., Phys. Rev. Letters 36 (1976) 703.

R.K. Carnegie et al., Phys. Letters 68B (1977) 287.
D.G.W.S. Leith, SLAC preprint SLAC-PUB-1980.
73) J.L. Rosner, in Proc. XVII Int. Conf. on High Energy Physics,
London (1974), ed. J.R. Smith, p. 11-171, Phys. Reports
llC (1974) 190.
75) G. Alexander, H.J. Lipkin and F. Scheck, Phys. Rev. Letters

17 (1966) 412.
76) E.M. Levin and L.L. Frankfurt, Zh. Eksp. i. Theor. Fiz-Pis'ma
Redakt 2 (1965) 105 (JETP Letters 2 (1965) 65).
77) S. Okubo, Phys. Rev. D13 (1976) 1994 and D14 (1976) 108.
78) F. Marzano et al., CERN Preprint EP/PHYS 77-5,

R.J. Hemingway, to be published in the Proc. 1977 Experimental
Meson Spectroscopy Conf., Northeaster University (1977).
79) E. Flaminio, J.D. Hansen, D.R.O. Morrison and N. Tovey,

Compilation of Cross-Sections. I. Proton induced reac-
tions. II. Antiproton induced reactions, CERN/HERA 70-2
and 70-3.
E. Bracci, J.P. Droulez, E. Flaminio, J.D. Hansen and
D.R.O. Morrison, Compilation of Cross-sections. (I and II)
CERN HERA 72-1 and 72-3.
J. Allaby, Yu. B. Bushnin, Yu.P. Gorin et al., Yad. Fiz. 12
(1970) 538; Phys. Letters 30B (1969) 500.
Yu.P. Gorin, S.P. Denisov, S.V. Donskov et al., Yad. Fiz 14
(1971); Phys. Letters 36B (1971) 415.
S.P. Denisov, S.V. Donskov, Yu.P. Gorin et al., Nuclear Phys.
B65 (1973) 1.
A.S. Carroll et al., Phys. Rev. Letters 33 (1974) 927 and
Phys. Letters 61B (1976) 303.
80) H.J. Lipkin, Phys. Rev. Dll (1975) 1827. There are some
errors in the values of the parameters quoted arising from
a confusion between several notations. The correct values
of the parameters actually used in this reference are
given with Eqs. (6.4) and (6.5) of the present paper.
See also, Nuclear Phys. B78 (1974) 381; Phys. Letters 56B
(1975) 76; High Energy Physics, Proc. of (1975) EPS Conf.,
Palermo, ed. A. Zichichi, Editrice Compositori Bologna
(1976) p. 1000, and FERMILAB-Pub-77/49-THY.
81) R.J. Eden, High Energy Collisions of Elementary Particles,

Cambridge University Press (1967) 1, 194.
82) U. Amaldi et al., Phys. Letters 66B (1977) 390.

136 H. LIPKIN
83) G.B. Yodh, to be published in the Proc. Brookhaven Symposium

on Prospects of Strong Interaction Physics at Isabelle,
April (1977). See also F. Siohan, Ph.D. Thesis, Univ.
of Maryland, PUblication No. PP776ll (1976) and T.K.
Gaisser, G.B. Yodh, V. Barger and F. Halzen, Proc. 14th
Int. Cosmic Ray Conference, Munich (1975), Vol 7, p. 2161.
DIS C U S S ION
CHAIRMAN: Prof. H.J. Lipkin

Scientific Secretaries: H. Bohr, A. Carter
DISCUSSION No.1
SCHNEIDER:
I have a question about your discussion of Han-Nambu
quarks, and the difficulty of distinguishing experiment-
ally between the Han-Nambu integer charge assignment and
the Gell-Mann-Zweig fractional charge assignments.
Could you explain this in more detail?
LIPKIN:
We are again running into the problem of reconciling
Quantum Mechanics with the parton model. Parton distri-
butions contain only intensity information, not phases,
and you will quickly run into trouble with QM if you
forget this fact. Let me give you an example. Consider
the case of the isoscalar EM current coupling to a pion.
This current has odd G-parity, so the final state X in
1{ lc:o...n:. ... ')( must have even G-pari ty. Now observe how
this conservation law is "violated" in the naive parton
picture. In that picture a quark absorbs the photon
with a probability given by F(x).
Probability N 'f()t) .
In this picture, the state which exists between the

absorption of the photon and the decay of the pion has
137
138 DISCUSSION
no memory of the G-parity of the photon, and any number

of pions can appear in the final state. Thus the parton
picture yields a conflict with a conservation law. What
is really going on? For each diagram in which a q absorbs
the photon, there is a diagram in which the G-parity
conjugate qabsorbs the photon. For each exclusive
decay channel, we must add these two amplitudes. For
channels containing an even number of pions in the final
state (G;+l), these amplitudes are in phase, and a non-
zero probability results. For channels containing an
odd number of pions in the final state, the amplitudes
are maximally out of phase and they cancel. You see,
therefore, that it is necessary to keep track of phases!
When trying to distinguish Han-Nambu quarks from the
usual fractionally charged quarks, you must realize that
in the deep inelastic parton picture we sum probabilities
for partons of various flavors to bear various charges,
averaged ~ color. Why averaged? A color singlet is
a coherent superposition of states in which a quark is
alternately red, blue and yellow. When describing a
process involving these quarks, we must calculate the
amplitudes for each color configuration and add the
amplitudes. If you do this for fractionally charged
quarks and for Han-Nambu quarks, you get the same answer.
The operator which differentiates the H-N EM current
from the fractional charged EM currents transforms as a
color octet. If physical states are color singlets, then
all the matrix elements of this operator vanish when
taken between the color singlet states.
COLEMAN:
Yes. In first order perturbation theory, there is

no way to distinguish H-N and fractional charge, since:
(singlet state \ octet operator \ singlet state) = 0
In second order, the octet operator can connect the
singlet state by passing through a color octet state.
But if color octet states are very heavy, or infinitely
heavy, then this higher order effect is killed by the
large energy denominators.
SCHNEIDER:
What can one say about the masses of the non-color-

singlet states?
DISCUSSION 139
LIPKIN:
Well, if you believe in the asymptotic freedom

picture, they are infinitely massive. That is the whole
point of confinement. If you do not believe in asymptotic
freedom, then all you can say is that they are heavier
than anything we have seen.
JEZABEK:
I am also puzzled by this. In the quark-parton model

there are sumrules which contain sums of squares of quark
charges, e.g.,
R = crle+e,---, h~("on$) =
(3" (ete: - P+\L" )
Since 2:Q is different for Han-Nambu quarks and for fract-

ionally charged quarks, one will obtain different answers
depending upon the scheme adopted. One could then decide,
in principle, which scheme is correct. What is wrong
with this argument?
LIPKIN:
The quark parton model is strictly valid only in

asymptopia, and since we are never there we always cheat
when we use formulas like ~ Q2. We never include the
b-quark in the sum, and we throwaway the charmed quark
below charm threshold. The assumption is that when we
are below the threshold for exciting a new degree of free-
dom, we can pretend that it does not exist. This is easy
when the threshold is for a new flavor. But it is much
more complicated below the threshold for a non-Abelian
quantum number, because it is not possible simply to throw
away certain terms. Phases are very important and you
must analyse the expressions for the transition amplitudes
including phases in order to see how to throwaway the
unphysical parts which describe excitations of states
below their threshold.
140 DISCUSSION
CHEN:
Is there any experiment one can perform in order to

distinguish the integral-charge Han-Nambu quark from the
fractionally charged quark? For instance, a Drell-Yan
qq annihilation process or the deep Compton scattering
process 1? ~ "( p +X ?
LIPKIN:
Any such experiment is performed over a time scale

much longer than the characteristic time defined by the
energy of excitation ~ E = E (color singlet) - E (color
octet) through Heisenberg's uncertainty principle. During
the time of measurement, the quark has had time to change
its color many times, and so you are only measuring the
charge of a color-averaged configuration.
MIETTINEN:
Professor Lipkin has emphasized in his lecture and

again in this discussion that in quantum mechanics it is
important to sum up amplitudes, not cross sections. He
also accused the parton model phenomenologists of being
careless in this respect. Just to show that not everyone
has forgotten quantum mechanics, I would like to mention
recent work by Brodsky, Carlson, and Suaya (SLAC-PUB-l79l,
published in Physical Review). These authors analyze the
charge asymmetry dO" CeTe,-_ h+y X) - dO" (e+e-~ h-yX' J
where the photon is detected at large Pt relative to the
hadron (~) direction. To lowest order in ~ the cross
section arises from interference between the amplitudes
(a) and the "Compton" amplitude (b).
y;Y /h
~x
(el)
~. X
(I
DISCUSSION 141
In the scaling region the hadron asymmetry is:
where the sum runs over all contributing quark fields

and the D's are standard quark fragmentation distributions.
We see that the asymmetry is proportional to the third
power of the quark charge.
WEILER:
I have a question which has been disturbing me for

a long time. When people calculate Zweig-violating decays,
such as those of the psi, they write down a diagram in
which the cc pair annihilates to a three-gluon inter-
mediate state (the lowest-order state of correct color
and charge-conjugation). These three gluons then material-
ize as a quark pair of a different flavor. Since six
fermion-gluon vertices appear, the process is proportional
to !C,t) ~ where the q2-dependence indicates that one
intends to use a "running coupling constant". My question
is: at what q2 should one evaluate the c~upling constant?
People usually say, "at q2:M~ ", but I do not see the
justification for this. Why not at q2:M2+ /3, or some-
thing else?
COLEMAN:
I think I might be able to clear this up. There are

really two questions involved here. The first one is not
understood by anyone, and that is simply that techniques
which have evolved from analysis of a space-like q2 region
cannot be applied with confidence to a situation in which
we are so clearly dealing with time-like q2.
As to the second question: by assuming the Euclidean
resul~s of QCD to apply to positive q2, what we learn is
that factorization, i.e., the result that the process is
proportional to ds(~~) .. 0(5 (q~) J occurs when only
vertex corrections are considered. Inclusion of the
complete, renormalization-group invariant set of diagrams
gives powers of Cls l'\1.) which are functions of a common
142 DISCUSSION
q2, i.e., ~~(~~). What we do not learn from QCD

is what mass scale to use for q2. It could be q2:M2f /3:
we simply do not know.
NILSSON:
You referred to the work of Arima and Iachello on

nuclear levels and showed some impressive predictions.
Is there any obvious way to see why they needed boson
constituents? And could you say what is unique in their
predictions compared with, for instance, the predictions
of the nuclear shell model?
LIPKIN:
A detailed answer would lead us into a long discussion

on nuclear physics. Let me therefore only remark that
the "bosons" are pairs of nucleons.
DINE:
Could you go into more detail on your interpretation

of the 2.8 GeV state at DESY as being the be? Have any
calculations been done in the bag model? And what then
becomes of the 't ?
o
LIPKIN:
The bag model people say there is some uncertainty

about putting four quarks in the bag and about charmed
quark interactions, so I think they would only have qual-
'c.
itative things to say on this question.
As for ~he it has many decay channels open with
small branch~ng rat~os, so it would not seem surprising
if it has not been seen yet. However, the Se interpret-
ation of the y events at DESY is an idea which has
occurred to me very recently, and there could be data
available which would contradict this interpretation. I
have not had time to look into this yet. Perhaps I will
discuss this in more detail in my next lecture.
(The following results from DASP on the X(2.8) were later

provided by Professor Lipkin. The source is a talk by
B. Wiik in the Wepion CERN School, June 1976.)
DISCUSSION 143
29 events above 14 background

branching ratio (l.5:t .4)xlO- 4
6 events above 1.3 background

branching ratio <: 3. 7xlO- S
at 90% confidence level.)
DINE:
Would the standard charmonium picture be in trouble
if your interpretation is correct?
LIPKIN:
No. In fact, It Hooft would like very much for the
'1., to be higher.
CELMASTER:
The processes ~ -''lY ) i' ~ yt'y I and ~ -. 'X(2..t.) "
is the 'c .
have been seen at DESY. You have suggested that X(2.8)
Judging from the fact that the first two
processes occur, we would guess that there is some mixing
of co pairs into the light mesons. Hence one might guess
that one could see also ~- b"f Why is this process
not seen?
LIPKIN:
The answer to your question is that if &c is a mole-
cule, it is difficult to say how much mixing will occur.
SANDA:
I have a comment about the possibility that X(2.8)
is a Sc' You mentioned in your lecture that -.!J'.... p Sc.
exists. Then ""' .. "( Sc also exists, and DESY should have
a good upper limit on this branching ratio since they
are very sensitive to a monoenergetic photon.*
* This point was discussed in Professor Lipkin's next

discussion section.
144 DISCUSSION
BALDINI-CELIO:
If "l' -'> 1'\ '( and '\J - rt'Y are larger than '-Y~ 1(0"(
because the former processes pass through a two-gluon
intermediate state and the latter through a three-gluon
intermediate state, as you have claimed, then why is
i"-Il "" so large?
LIPKIN:
It is all the same argument. The diagram
can be either considered as_part of a transition matrix

element or as_a mixing of cc into the 'Y'l. and rtf or as
a mixing of qq into the ~ You can take either point
of view and the result is t~e same to first order in
this diagram.
The process is represented by
You can either include the cc ~qq as part of the '1. wave
function and say that it is mixed, or you can say that
the transition is a two-step process ""/~'tY'/C-'t'1.
The result is the same to first order in the OZI-
violating diagram.
BALDINI-CELIO:
Why do people not .

cons~der the ...,'
-L exotic, if its
width is so small?
LIPKIN:
The allowed decay tt' ~ rt 7t 7[' is small simply

because phase space is limited.
DISCUSSION 145
DISCUSSION No.2 (Scientific Secretary: R.V. Budny)
DEO:
Please outline the proof of Rosner's theorem that

states that duality and baryons imply BE exotics.
LIPKIN:
Rosner showed that the standard, finite energy sum

rules used to get the nice duality results for meson-meson
scattering lead to trouble for baryon-baryon scattering
unless there are exotics. The simplest way to see this
is to consider A-A scattering:
ft+----r-- A+ .+ 6+
f)A,. p,A a
A
-0
_---'L..-- -+
~- tf
!:::.
M1 M20
In the Regge region, these involve p and A2 exchange,

which differ by signature and thus have opposite sign
contributions. We can extract Im(Ml-M 2 )= p-exchange,
which has a finite contribution unless the coupling pAA
vanishes, but we do not believe it does. Similarly
Im(Ml+M2)=A2 exchange. Duality implies that there is a
resonance in the AA system. Notice J::t+6.0 has charge
2 and A+~A- is exotic. We can conclude that all these
couplings vanish or else there is something other than a
resonance contributing which would violate simple
"resonance-saturation" duality. Rosner concluded that
the p Ab. coupling does not vanish and that duality is
correct so there must be resonances.
COLEMAN :
I did not understand your group theoretic argument

for two couplings.
146 DISCUSSION
I can write
This means that your conclusions about two couplings are

confusing.
LIPKIN:
The idea is that if a baryon-anti-baryon annihilate,
you can get two quarks coupled as a vector and two anti-
quarks coupled as a vector. The question is: can it
decay to two scalars?
-=--
v
" 1/ .=0
f. S
-
0 ~
0 - II +
#~ 0 o=-e s
0=0
v
You can think of this in terms of wave functions which

are expanded in different bases. One piece can come
apart in the horizontal direction, but not in the vertical
direction. The situation is reversed for the other piece.
Thus the breakup is complicated.
It is not obvious whether it is easier for the state
to create a pair and break up into a BS or to recombine
into two mesons.
JEZABEK:
Is it possible to have Han-Nambu quarks with color

being an exact symmetry?
LIPKIN:
With Han-Nambu quarks, the electromagnetic current
cannot be a color singlet so color cannot be an exact
symmetry. There must be high mass states which are not
color singlets.
DISCUSSION 147
JEZABEK:
Could you comment why the AI=~ rule supports the

existence of color?
LIPKIN:
There is an old derivation of the AI=~ rule using

the Iour-fermion interaction and Fierz reshuffling which
requires symmetric instead of anti-symmetric quark
statistics. This is reviewed by O.W. Greenberg and
C.A. Nelson (in press in Physics Reports).
MESS:
I would like to remind the audience that there are

data from the DASP group on the decay ,\,'.,. '1)( - '("'f "(. The
result is a limit B.R. ~ 3>."'~ 10- 3 (90% c.l.) based on 6
events in the 2.7 to 2.9 mass region compared with a 1.3
event background; compare this with the result
:r/I.j1~ y'X- 'f'1--r, with B.R. ~ (p;O.4-) 'It. '10-+. More data
is needed for ~'-" '("fY .
LIPKIN:
This means that 'l" .... y'( y is not excluded because

the limit is 20 times larger than the branching ratio
.T/~ --. '{'('( via X.
TOWNSEND:
You showed that for 4-quark states, you do not get

exotic quantum numbers due to flavor antisymmetry. What
if we reverse the argument: if we find exotics, presumably
it is evidence for 4 flavors. If we observe 6 quark
exotics, is it evidence for 6 flavors?
LIPKIN:
Only the lowest states should be flavor anti-

symmetric. Jaffe predicts higher states which are not
flavor antisymmetric. For 6 quarks, flavor antisymmetry
suggests uuddss states.
148 DISCUSSION
MORINESCU:
The flavor symmetry is usually broken by giving diff-

erent masses to different flavored quarks. How can quarks
with different flavors be considered identical particles
which are anti symmetric?
COLEMAN:
In a theory with many kinds of fermions, the general-

ized Pauli principle results from assuming that fermion
fields anticommute.
KOCKER:
I would like to comment on your assumption that the

~- has spin 3/2. Until now, there was no experimental
evidence for the spin of the.n-: At the Budapest confer-
ence, the ABCLV collaboration presented the first strong
evidence that the spin is greater than ~.
MIETTINEN:
Can you explain why baryonium resonances are so

narrow?
LIPKIN:
If they have high spin, the centrifugal barrier might

prevent them from decaying. In the M.I.T. bag model, the
bag would be very deformed for high spin.
MIETTINEN:
Chew and Jaffe seem to have very different interpre-

tations of QQgQ states. Chew expects them to couple
strongly to BB, and weakly or not at all to mesons, where-
as Jaffe expects them to couple strongly to mesonic chan-
nels and practically fall apart.
LIPKIN:
The Chew and Rosenzweig m~del has duality built into

it, so their QQQQ states are BB resonances. I do not know,
DISCUSSION 149
off hand, what the spin of these states are predicted to

be. Jaffe's model predicts that QQQQ with no orbital
angular momentum can fall apart into two s state mesons
immediately, and will be very wide. For the configuration
with high orbital angular momentum:
there can be less energy in the state where QQ and separ-

ately QQ are antisymmetrized and then coupled together
with high orbital angular momentum. It may also decay
preferentially to BB. The real question is whether
Veneziano or Chew and Rosenzweig predi2t that low spin
states would decay preferentially to BB; then they would
disagree with Jaffe's prediction. For high spin states,
they are saying the same thing, even though they come to
their conclusions from very different paths.
MIETTINEN:
I wish to point out that what you say indicates that

it is very interesting to measure the spins of these
objects.
DINE:
In arguing that diquark interactions are repulsive

and diquark states are unstable, you considered meson
exchange graphs. Are there any arguments which generalize
this to higher order?
LIPKIN:
As far as I know the argument is based just on the

lowest order graphs.
COLEMAN:
At the risk of exciting universal contempt, I will

state that it is true in two dimensions.
150 DISCUSSION
BERLAD:
Why is the duality diagram
& :;? M
~ ~M
forbidden by the Zweig rule?
LIPKIN:
For a 3-quark vertex, the Zweig rule is well defined:
is allowed, but
is forbidden.
t ___
~ ---..
One can define this selection rule ~n either of two ways:
1) hairpin diagrams (disconnected qq pairs of the same
flavor) are suppressed; 2) diagrams with disconnected
parts are suppressed. For 3-quarks, both give the same
Zweig rule, but with 4-quarks, there are ambiguities.
For instance
::::>,,1/
.
is forbidden by both definitions. The graph
,\,1
~2) ~: (-.v'...., ~ nTC )
is disconnected, but not hairpin-like. Moreover, if you

cross it, you get the Pomeron (n-i'-.1t-.!J) so in this case
it is not clear whether the Zweig rule holds.
DISCUSSION 151
The case of baryonium,
--~
---~-
---
is prohibited by Rosner, but allowed by Jaffe. Also
V~
~Tr
presumably is double Zweig forbidden. My conclusion is
that there is no dynamical understanding of the Zweig
rule, so the generalization to the 4-quark case is not
unique, and may be impossible.
BIZZARRI:
Does your argument that molecular forces cannot bind

mesons also apply to two baryons?
LIPKIN:
For the case of two mesons, we could try to arrange

the Q and Q farther apart than the Q and Q pairs. For a
sharp square well potential, we could attempt to bind in
the configuration
Q
Q
where the two Q's are outside the range of the potential,
but the two QQ pairs are inside. Although this might
work, it is unrealistic to assume that the potential has
sufficiently sharp sides. In the more realistic cases
of smooth sided potentials, it does not work.
COLEI1AN:
Doesn't the argument boil down to just this: what we

mean by molecular forces is instantaneous action at a
distance between two charge densities. Now there is a
big difference between electromagnetism and a gauge theory.
The electromagnetic charge density commutes with charge,
therefore it can have a non-zero matrix element between
neutral states, however the color charge density
152 DISCUSSION
transforms as an octet under the color group, and thus

cannot have a non-zero matrix element between color
singlet states. Therefore you can never get anything
analogous to molecular forces.
LIPKIN:
This is true. The four quark configuration is not
strictly a molecule but more like a microcrystal, so
you might still expect binding. But even though the
four quarks are not arranged into two color singlet
QQ pairs, it still does not work.
DISCUSSION No.3 (Scientific Secretary: M. Dine)
BERLAD:
Can you show us the five parameters fit to the total
cross section you showed us this morning and explain the
two new terms? Could you also explain how you obtain the
real part?
LIPKIN:
The fit is to
where
CT2 (Hp) -= Net N ns (?I"b/20) -'

CJR (I-Ip') = l ~ m+ 2 Nt ) ( t>ralt /20 ~-%.
Cf = b.S Mb
C2. = 2..'2. mb
c~ = 1.15 mb
t = C.I:' ,) S :.0.2.
DISCUSSION 153
I should recall at this point that one of the "whys"

I was to talk about was: "Why do single quark transitions
dominate; why is it so easy to have spectator quarks?"
The additive quark model works much better than it ever
was supposed to. All kinds of reactions are decribed by
saying that in any transition there is one active quark
which makes the transitions while all the others are
spectators.
I discussed examples in the case of pseudoscalar
meson production and in other contexts. Why this picture
should work for high mass resonances is the kind of quest-
ion we have asked for a long time. I started to work on
this problem because I felt it was time to look for proc-
esses in which the single quark picture breaks down.
Total cross sections seemed the obvious place to look,
for these cross sections are supposed to satisfy the 3/2
rule for <IN ..... /<T1TN This is broken at the 15% level.
As better data became available, it was of interest to
try and describe this deviation by two quark processes.
This deviation occurs in the isoscalar exchange
amplitude. If this amplitude consists of two exchanges,
one is a Pomeron and the other is an isoscalar, even
signature exchange. The obvious exchange to try is that
of an f. Consider the simplest description of total cross
sections, a Regge term given exactly by duality diagrams
and a Pomeron term which simply counts quarks and gives
the 3/2 rule, and a third term which transforms like the
product of the Pomeron and the f (the f couples only to
non-strange quarks). With data for C17t'i' p ) cr~~ p ) CJ~ P ClYId <:f"pp
there are more cross sections than independent contribu-
tions and the simple model can be compared with experi-
ment. It works strikingly well, and the new data (the
Fermilab data came after this model) continues to fit
the model. One can see from the slowly decreasing energy
dependence that it cannot be a Pomeron-f cut, as one
might expect for such a double exchange. It might be a
triple Regge term involving a Pomeron, an f, and another
Pomeron. There have been suggestions and models with
the Pomeron containing two pieces.
BERLAD:
If you have e ') 0 aren't you in trouble with

the Froissart bound?
154 DISCUSSION
LIPKIN:
Of course. If you go to high enough energies this
model won't work. But it is the simplest thing to do;
anything better than a power would involve more parameters
Getting the real part is trivial, because if you have
an analytic expression involving powers you get an express
ion for the real part by putting in the standard phases.
MARTIN:
As long as you use this expression far from the regio
where it breaks down, there is no problem with getting the
real part.
BERLAD:
But if you use this expression you may need a second
subtraction, whereas the usual expression has one sub-
traction.
MARTIN:
Yes, but for an order of magnitude estimate it is OK.
WEILER:
Two conclusions from this morning are: first, the ~I
is heavy; second, the ~ has components other than SU(3)
octet and singlet. Can both of these problems be approach
ed by mixing in some cc? Does this conflict with known
-:r/~ branching ratios?
LIPKIN:
The problem of the ~' being too heavy is not solved
by mixing in something heavier. The wave function for the
lowest state having a given set of quantum numbers can onl
be pushed down (variational principle). From the point of
view of perturbation theory, the off diagonal piece
dominates over the additional mass term.
DISCUSSION 155
DINE:
Do you have any ideas for a mechanism to push the

state up?
LIPKIN:
I thought the instanton might have something to do

with it, but Sidney Coleman says that the instanton only
tells you that the \/ is not a Goldstone boson; it says
nothing about its mass.
' t Hooft has suggested the possibility of mixing
with a lower energy state of pure glue. But we see no
pseudoscalar SU(3) meson lower than this; so I don't
know how this would work.
HINCHLIFFE:
Can you comment on the possible existence of

glueballs?
LIPKIN:
I can only say that it is possible that they are

there, and that it is possible that they mix with ordinary
mesons. It is possible that if there is an extra piece
in the ~' which does not contribute to these sum rules
it might be a glueball state rather than a radial
excitation.
BALDINI:
Let me insist on a question I asked you already. You

have pointed out that the ~ is a puzzling particle. The
width of this particle is as yet unknown, but the upper
limit seems incredibly small (800 keY). What is the m~n~
mum value below which you would begin to worry about this
width?
LIPKIN:
The only allowed hadronic decay of the ~' is to ~~~.

The other decays are electromagnet ic t yt'- py). I think
that the narrow width is just a question of phase space,
156 DISCUSSION
so I see no cause for concern. I would have to do a

calculation to determine a minimum width.
With regard to the question of ~ .... rt'y and mixing
let me say the following:
The usual description of this process is that the ~

emits a photon, ~ecoming a 0- state which annihilates,
giving rise to qq. Since the mixing is small, you can
say that the rt is already mixed into the 'l. l~) wave
function, and g~t the same result. The interesting point
is that the diagram
where the cc first annihilates and the photon is then

emi tted allows you to make a 1t0 (the other will not,
since you have isospin zero). Since this process is much
weaker, the annihilation must go faster in the 0- than
in the 1- state.
But Rosenweig points out that these two processes
take place at different masses, so that the difference
might reflect the mass dependence of the diagram.
MARTIN:
I have several comments on the pp cross sections. I
noticed (as did Arthur Wightman) that the highest energy
points of the CERN-Rome experiment sit below their best
fit. According to Ugo Amaldi this reflects an energy-
independent uncertainty in their overall normalization.
This normalization is left free, within reasonable limits,
in their fit.
DISCUSSION 157
The fact that their seven parameters fit indicates

rising cross sections shows that even with a flexible fit
you are forced to have rising cross sections well beyond
the last point at which the cross sections have been
measured.
The fact that Harry Lipkin uses powers in his fit
does not worry me, for we are far from the asymptotic
regime in any theoretical model (Cheng-Wu, ter Martirosyan,
Le Bellac, et al.) and in practice. For instance if one
calculates the real part from a fit of tot to A + B-
(log s)2 one finds that f~ the real part, reaches its
asymptotic form ) ~/log s predicted by Khuri and
Kinoshi ta very slowly. In fact p reaches its maximum
at about 10 5 GeV, i.e. around Isabelle energies. Also,
practically all models predict a black disk asymptotically,
Le. ()I(l~t-i~ ~(i'totand we are very far from that.
I would also like to say that information about pp
cross sections is very important, and I hope that anti-
protons in the SPS will be used to measure total cross
sections.
LIPKIN:
First, my point was not that my five parameters fit

was better than CERN's seven parameters fit, but simply
that fits to complicated models are not very meaningfuL
There is just not enough information in the data.
If it were trivial to get the real part from the
imaginary part, there would be no information in the fact
that the real part fits the data. Getting the real part
successfully from the formula indicates that the formula
will continue to fit the data at higher energy. This is
consistent with the information we have from the cosmic
ray data.
There are, however, things which can be learned from
better cross section measurements. There seems to be a
slight discrepancy at the highest Fermilab energy. In the
plot on a log-log scale of the cJ"Tlt-l-eJ"l'ldifference and the
difference between Y3 eJp P and 1~ cr~+p
(the deviation from the 3/2 ratio of baryon to meson cross
sections), <T"'1fN - G""Jo'.t-.l appears to level off and remain
constant, whereas the deviation from the quark model form-
ula appears to be decreasing. So it appears that ~/s
and K's will continue to look different at high energies,
whereas the predictions of the quark model will be
158 DISCUSSION
fulfilled. This will, of course, disagree with my formula

but the basic physics questions are more important. A few
more points at 4-500 GeV would show if these patterns
continue. So there is a point to measuring total cross
sections for hadrons other than protons in the intermed-
iate range.
One comment on pp. There is another puzzle of inter-
est. Certain quark model relations for total cross sect-
ions work much better than one would expect. One is the
GO universality relation which relates the difference
between OPP and ()\,p to meson-baryon cross sections.
If you use the optical theorem, the difference between
pp and pp is just ~ exchange (apart from a small correct-
ion for .p exchange which you can include). The G:>
coupling counts non-strange quarks, so the ~ coupling
to the proton is three times the CA) coupling to the K,
so the C:S-,,~_(f"pp difference is three times the CJ"""+p - 0-"-1>
difference, and if you throw in the (S"",+p -(f"7t-" data you
take care of the p exchange. This relation fits remark-
ably well all the way from 6 GeV to the highest Fermilab
energies. If you turn the optical theorem around and
recognize that you have successfully used the simple quark
model to relate very different processes, you have a great
mystery, which one would like to understand more deeply.
DO MESONS FILL SU(3) NONETS?
R. Bizzarri
Istituto di Fisica, Universita di Roma
Istituto Nazionale di Fisica Nucleare, Sezione di Roma
1. INTRODUCTION
One of the greatest successes of the SU(3) symmetry and the

naive quark model is its ability to classify the observed spectrum
of particle states. The rules are particularly simple in the case
of mesons: a qq pair in a state of angular mome,ntum L and spin S
(S 0,1) can give rise to mesons with quantum numbers
P
C
(_l)L+l
(-1) L+S
Le., J L for S
-+ -+ -+
J L-l, L, L + 1 for S 1 .
J L + S
Furthermore, since the quarks have three different flavours*), there

are 3 x 3 = 9 different states for each value of J PC .
For an exact SU(3) symmetry these 9 states should break into

an octet and a singlet. In terms of isotopic spin the octet is
made of an I 1, strangeness s = triplet, two I = %, s = l
doublets and one I = 0, s = singlet. This last state, due to
SU(3) breaking, will mix with the SU(3) singlet having the same
*) "charm" will be ignored in these lectures.
159
160 R. BIZZARRI
quantum numbers, and therefore we shall talk of SU(3) nonets with

two I = 0 s = 0 members.
If one considers the lowest angular momenta (L < 2) the pre-

dicted.
A dotted line underlines the states which cannot be considered

well-established members of the nonets either because their quantum
numbers are not established or because their existence is still in
doubt. Table 1 shows that the two L = 0 nonets are well estab-
lished: a great success of SU(3). The situation for the L = 2
nonets is still very preliminary since only recently this mass
region began to be investigated with sufficient accuracy. However,
the situation is not much better for the L = 1 nonets which are
being studied since a long time. It is interesting to compare this
Table 1
J PC
iI L S
--
I = 1 1/2 0 0
0 1 1 p K* (892) w
0 0-+ K (958)
11 11 11'
2++ K* (1420) f (1270) f' (1514)

1 1 A2 (1310)
1++
Al..
Q
..A . .. . . . .........
. .D. .(1285) E (1420)
0++
P (970)
..
K
... E: ~:..(??.3).
+-
0 1 B (1235) QB - -
--
..
2 1 3 g (1680) (1800) w (1675) -
-- ~........
2 - - - -
--
1
p'.........
(1600) - - -
+-
0 2 (1640) L (1770) - -
A3......... ........
DO MESONS FILL SU(3) NONETS? 161
Table 2
Particle Mass Particle Hass
K1Kl(S*) 1000 KK TI (E) 1410
TI W (B) 1220 K TITI (Q) I 1175
TI P (A l ) 1080 K 730
TI p (A 2 ) 1310 2 TI (<:) 381
Table with Table 2, taken from the lectures by S.M. Berman at the
1964 School of Erice (with present day names added) which showed
the observed "meson resonances which at the present time are not
fitted into some SU(3) multiplet structure"l). Of these states the
++
A2 is a well-established member of the 2 nonet, the B is a well-
established 1+- state but still without a nonet. The existence of
an S* pole is still questioned, strong doubts exist on the resonant
nature of the Al , the quantum numbers of the E meson are not estab-
lished. As for the S-wave TITI and TIK resonances (<: and K) their
presumed mass has been steadily moving upwards and only recently
began to stabilize around 1300 MeV. The fact that 13 years have
not been sufficient to clarify the situation is somewhat worrying
and might be taken as an indication that something does not work ~n
the naive quark model. In my lectures I shall not discuss the

healthy 2++ nonet but focus my attention on the L = 1, J PC = 0++
++ +- .
1 , 1 nonets 1n an attempt to give a feeling of the difficulties
encountered in their study. My personal conclusion will be that, op-
timistically, these nonets are there and the states for which there
is experimental indication could be genuine resonances. I shall not
attempt in my discussion to be complete or fully up-to-date.
162 R. BIZZARRI
2. The 0+ Nonet
2.1 I = 1
The I = 1 member of this nonet could be identified with the

0(980) meson which has been observed in production experiments as
an enhancement in the nTI combined mass. As an example I shall dis-
cuss recent data obtained by the Amsterdam-CERN-Nijmegen-Oxford
(ACNO) collaboration 2 ) from an 80 events/~b sample of Kp interac-
tions at 4.2 GeV/c.
Starting with 16271 events fitting the reaction

\e\=>--,> /\0 llTn,+ll-x-n0
a sample of 245 events has been selected representative of the re-

action
"-P-' L.+t.1?>~S) 1. 1l-
~ /\0n. + L. l lTIt-lt 0
with a peripheral cut t' = It I - It m1n
. I < 1 GeV2, t being the
squared four momentum transfer from the target proton to the (ATI+)
system. The resulting nTI mass spectrum shows a conspicuous peak
near 1 GeV (Fig. la). A fit with an S-wave Breit-Wigner parametri-
zation gives
The observed enhancement is near to the KK threshold (991 MeV)

and SU3 predicts a sizable coupling to the KK system. In fact, an
enhancement has been observed in the charged KK system at threshold
which can be associated with the 0 meson (for this experiment see
Fig. lb).
In this situation the use of a simple Breit-Wigner parametri-

zation for the observed nTI peak is not adequate. A more correct
parametrization could be
:z.
dcr,,'1. =
dm (1)
with (for an S-wave resonance)

25
>
GI
(!) 20
('II
o
o
......
15
en
~
Z
UJ
>
UJ 10
o
UJ
~
X
(!)
UJ
~
0 .6 0.8
Fig. la nn mass distributions for events of reaction

K-p ~ An+n-n weighted by the probability of
events belonging to the reaction K-p ~
~ I+(1385)nn. The dashed curve is the expected
distribution without inclusion of 8(970). The
full curve has included the 8(970) contribution.
10
~
(!)
N
o
..g
en 5
~
z
UJ
>
UJ
o.~6----~-----A~~--L-----~~WC----~
0.8 1.0 1.2 1,4 1.6
M (K --;KO) , GeV
Fig. lb KOK- mass distributions from the reaction

K-p ~ I+(1385) KOK-. The full curve represents
the KK threshold contribution from the 8- plus
phase space.
164 R. BIZZARRI
above KK threshold
below KK threshold
where qn is the 11 momentum in the llTI system, ~ and g.!] may be con-
sidered as squares of the coupling constants to the KK and llTI sys-
tems, fo is a measure of the coupling to the initial state, it may
be considered constant ~n a production experiment.
It is interesting to discuss the properties of formula (1).

Above KK threshold it gives
~~ = C. M~ ro r 1t y\/ lm~-m~)2-\- m~ \q\C.J~-~ ~ \ltyt):&

and the ~ term in the denominator will cause a rapid decrease of
the cross-section with increasing m, almost irrespectively of the
value of f TIll , provided ~/gll is sufficiently large. Below threshold
2. \' n\
c.. m~ 0 1t'1
and the ~ term in the denominator will cause a rapid decrease of

the cross-section for decreasing m, below~. The result is that
the llTI mass spectrum will peak at the KK threshold and have a small
width for a wide range of values of ITL and f , provided the ratio
K TIll
9K/ gll is suitably chosen.
This effect has been explicitly shown by Flatte 3) who was

able to fit the ACNO data with values of ~ between 800 and 1000
MeV and width up to 300 MeV (see Fig. 2). The point is that the
experimentally measured mass and width of the llTI peak are not suf-
ficient to determine the three parameters ~, gil' 9K present in
the formula (1). SU(3) gives a firm prediction on the ratio
9K/ gll = 3/2. Assuming this ratio the uncertainty in the mass is
largely removed and a fit to the ACNO data with formula (1) gives
., bl
"' ~ _ ...v ""R::' eoo ".-"I
0 :;: 82 M.V r 0 ;; 3C() III_V
C : 91 J'b/GoV C = ~,.IjIG .v
Iv., ' 2.0 g.;g ~ : 4.0
O ~~~~~~~~~~LL~~
80
920 1040
MASS OF DECAYING SYSTEM (MeV)
Fig. 2
-
KK and TIn mass distributions from Ref. 2. The
curves represent the theoretical distributions
obtained from a two-channel resonance with
parameter values given in the figures.
166 R. BIZZARRI
(full line curve in Fig. 1).
2.2 I =0
The I = 0 members of this nonet are expected to be strongly
coupled to the TITI and KK systems. The TITI interaction is not direc-
tly accessible to experiment but it can be studied by interaction
of pions on virtual pions with reactions like
-t A++ (2)
"It. p~ u n+7t-
R- t> ~ h R+TC (3)
which at sufficiently low It I are expected to be dominated by one

pion exchange (Fig. 3). In fact if the cross-section can be extra-
polated to the unphysical region at t = ~2, the equality holds (for
instance for reaction (3)):
G'2. -t:
=
47t.
with: ~ pion mass

t four momentum transfer between the target proton and
the outgoing nucleon
qTI momentum of the TI in the dipion c.m.s.
q initial momentum in the over-all c.m.s.
i
s = total energy in the over-all c.m.s.
The relevant variables in this extrapolation are the TITI scat-

tering angle cos 8*, Hand t. If the data were precise enough to
TITI
allow an unambiguous extrapolation to t = ~2 that would be enough.
However this is not the case and the extrapolation must be guided
by some assumptions as to the t-dependence of the production ampli-
n_ __n
----Y:-=--=--=---n
n: I
I
N
Fig. 3 The one-pion exchange contribution
to dipion production.
3roOO ~----~t~----------------'
++
.......
>- +
II>
+ +
-
~
o +
N
.' + +
en
I- t .
+
z
UJ t f
>
UJ + +
'.J
O ~~~~--~----~----~-------J
0.2 0.6 1.0 1.4 1.8 21
+ - - +
Fig. 4 TI TImass spectrum for events TI p ~ TI TI n
with It I < 0.15 GeV/c. Histogram: ob-
served events. Dots: weighted observed
events. Points with error bars: pro-
duced events.
168 R. BIZZARRI
tudes. This requires an amplitude analysis of the process in the

physical region as function of the four independent kinematical
variables and in terms of independent helicity amplitudes. A task
which can be performed only making some assumptions based on a
judicious choice of theoretical prejudices. The values of the
extrapolated cross-sections obtained with different but reasonable
descriptions of the physical amplitudes agree quite closely with
each-other and the resulting TITI phase-shifts can be considered quite
well established up to a value of M 1300 MeV.
TITI
I shall ignore the large number of very important contributions

to the study of this reaction and only briefly discuss the results
from the most recent high statistics data obtained by the CERN-
Munich collaboration 4 ). Reaction (3), induced by 17 GeV/c incident
pions, was studied with a forward spectrometer capable of identify-
ing the two charged pions and measuring their momentum. The ob-
served TITI mass spectrum for JtJ< 1.5 (GeV/c)2 is shown by the his-
togram ~n Fig. 4. The dots indicate the spectrum when corrected
for experimental inefficiencies. Since the forward spectrometer
covers only a limited solid angle, the observed angular distribu-
tions of the dipions have been fitted with the appropriate number
of spherical harmonics. Integration to the whole solid angle then
gives the number of the produced events indicated by the points
with error bars. I shall not enter into the details of the extra-
polation to the pion pole and the phase-shift analysis. The phase-
shifts o~ and inelasticities n~ can be unambiguously determined at
least up to ~ 1300 MeV (above this energy ambiguities arise) and
are shown for S, P, D waves in Fig. 5 (Ref. 5) and the correspon-
ding Argand diagrams in Fig. 6. There is a very clear evidence for
the p, f and g resonances which have a n~ce Breit-Wigner behaviour
with a different degree of inelasticity (see also Fig. 7). There
is an indication of a second inelastic P-wave resonance at 1610 MeV
which however is not present in other phase-shift analysis. As
for the s-wave TITI phase-shift and inelasticity, they are shown in
~O'r-----------------------------------~
1.0
U5 1):
a
180"
150
,
{j 0
120
9(t
SOo
30
1.0
Q5 rf,
a
10 ~
0 ---
170
1.0
0.5 ,
1)'
o.6 1.6 GeV 1.8

I I
Fig. 5 phase shifts oJ and inelasticities nJ for
TITI
P, D, and F waves from energy dependent
(points) and energy independent (curves)
fits.
170 R. BIZZARRI
Fig. 6 Argand diagrams for the TITI partial wave

amplitudes from an energy-dependent fit.
Fig. 7 n+n- intensities: a) I (1 = 1); b) I (1 = 2);

c) I (1 = 3) in the ,f, and g resonance re-
gions. The curves are Breit-Wigner fits.
172 R. BIZZARRI
Figure 8 and the corresponding Argand diagram in Fig. 6. The

behaviour of this amplitude ~s very peculiar and, although the
phase-shift crosses 90 0 and 270 0 , it does not resemble at all a
Breit-Wigner. The most striking phenomenon is the sudden drop of
the elasticity at the KK threshold associated with a very rapid
variation of the amplitude phase which jumps from 100 0 to 240 0 in
a mass interval of less than 100 HeV. To interpret this behaviour
the most straightforward possibility is to assume t~is rapid jump
to be due to an inelastic resonance (the S*) occurring at the KK
threshold on a background with a very large phase-shift.
The determination of the parameters of this resonance ~s not

straightforward but the estimates agree on values like
-tt
m
s"*'
~ 'l000 MeV rS = 60+200 MeV
- S* S*
and its coupling to the KK system must be very large: g-- > g--
KK TITI
Therefore this experimentally narrow effect is again associated
to a resonance which can be considered "wide" ir: the sense that its
-
observed width is limited by the nearness of the KK threshold and
not due to a small value of the coupling constants.
The question is now: is the S* the only S-wave TITI resonance?

To answer this question one can subtract to the og the effect of
the st, and look at the residual phase shift (Horgan, Ref. 6).
~0
o
= bO0 _ \ b: ') '0,"'"
3600
3300
3000
2400
1.0
0.5 1JO
D
o
6
Fig. 8 ng
The I = 0 s-wave TITI phase-shift 8~ and inelasticity
as obtained from energy independent (points) ano energy
dependent (curves) fits.
174 R. BIZZARRI
The result is shown In Fig. 9 (dotted line). The residual phase-

shift is seen to go slowly through 90 0 at a mass m ~ 1200 MeV.
This can be interpreted as a very wide (r E '" 600 MeV), mostly
nn
elastic, resonance. Again exact values of the mass and width cannot
be determined due to the large width and the need to subtrace the
S* effect. This interpretation of o~ is not unique and has been
questioned?). It has however the great quality of being quite
. ++
straightforward and gives the two expected 0 I = 0 resonances.
2.3 I = !z
The I = !z S = l members of this nonet are expected to couple
to the S-wave Kn system which can be studied from the reactions
K P -+ K n+ n (4)
K p -+ K n- 6++ (5)
by the technique of extrapolation to the pion pole already discus-

sed. I shall limit myself to mentioning the results of a phase-
shift analysis 8 ) of the recent results from a 13 GeV/c SLAC spec-
trometer study 9) of reaction (4) and (5). The phase-shift solution
is unambiguous up to 1.48 GeV but there are four possible solutions
above this mass. Below 1.1 GeV the important I = !z waves are the
S-wave slowly rising up to ~ 50 0 and the P-wave going through the
elastic K* (890) resonance. The Argand diagrams for the S, P, D,
F waves above 1.1 GeV are shown in Fig. 10.
In all four solutions the S-wave describes a resonant circle

with a mass (defined as the point of maximum speed on the Argand
diagram) mK ~ 1400 -;- 1450 MeV, width r~ = 225 -;- 325 MeV and elas-
ticity x = 0.5 -;- 0.9.
2.4 Conclusion
The most recent data suggest the existence of a full 0++ nonet
made of resonances with a very large width.
150
d~
(degrees)
120
90
60
0.2 1.6
Fig. 9 The experimentally determined 68 plotted module

180 0 (full line) and the residual 88 after sub-
traction of the S* contribution (dotted line).
176 R. BIZZARRI
(8,3 03
SOLUTION A SOLUTION 8 SOLUTION c SOLUTION 0
'3
1.1 1,1 1,1
( )
S' 1.5 S'
1.5
1.7 1.7
1.7 ,5
M,.5 /
~~'3 P,,-~
1.7
1,5 1,3
~F~~F~ 1,5 1,5 1,5
Fig. 10 Argand diagrams for the four solutions for KTI partial waves.
The masses of a nonet should satisfy the Gell-Mann Okubo for-

mula
which requires at least one of the non-strange mesons to have a

mass higher than the strange meson. Present data suggest mE ~
1300 MeV and m ~ 1400 MeV but the uncertainties are large and
K
m > m cannot be excluded. In this respect it should also be
E K
noted that a clear and practical definition of the resonance mass
in the case of wide resonances in the presence of large background
is lacking. Certainly the mixing for this nonet is far from ideal
and the mixing angle 6 should be quite large.
3. 1+ Nonets
3.1 I = 1
The 1+- I = 1 particle is the well-established B(1235) meson

which decays into WTI. The 1++ particle should have G-parity =-
and decay into pTI. A natural candidate has been the enhancement on
the pTI mass at 1100 MeV known since as long a time as AI' abun-
dantly produced in diffractive TIN reactions. I shall limit myself
to a brief discussion of the results of the CERN-IHEP Boson Spec-
trometer 10 ). The reaction studied ~s
TI P -+ TI - TI - TI +P (6)
at 25 and 40 GeV/c. The slow recoil proton is analyzed in time of

flight and direction in a proton telescope and at least two of the
three forward tracks are momentum analyzed in a magnetic spectro-
meter.
The 3TI mass spectra for different cuts on the four-momentum

transfer squared t to the proton are shown in Fig. 11. The events
with 0.17 < It I < 0.33 (GeV/c)2 show a shoulder (the AI) on the low
mass side of a prominent A2 peak. For the low It I events (0.04 <
It I < 0.17 (GeV/c)2), the Al appears as an enhancement with mass

178 R. BIZZARRI
mltf ALL EVENTS

:>
II)
TTTTT1I OI m 665<mn-:r-<865MeV
~
C
--z
N
(I) 250 a) 0.04<ltI<0.17 !GeV/d
~
~
:>
~
2.5
b) 0.17<ItI<o.33 IGeV/d
:>
II)
~
C
--
N
500
~
z
~
:>
~
1.5 2.0 2.5

msnLG eVJ
Fig. 11 3n mass distributions from reaction n-p ~ n+n n p
at 40 GeV/c corrected for acceptance, (a) in the
interval 0.04 < It I < 0.17 (GeV/c)2, (b) in the
interval 0.17 < t < 0.33 (GeV/c)2. Events with
0.665 < mn +n-< 0.865 GeV (p band) are dashed.
'\, 1150 HeV and width '\, 300 HeV. To deduce the properties of this
3TI system a partial wave analysis is performed. The cross-section
of reaction (6) is written
(7)
where s total c.m. energy squared

t four-momentum transfer squared to the photon
m3TI mass of the 3TI system
are the production variables for the 3TI system, its decay variables
being
energies of TI+TI- subsystems, i.e. coordinates of the

3TI Dalitz plot
a,S,y angular coordinates to specify the orientation of
the decay particles.
Pab is the density matrix for the production of a 3TI state. Such
a state is assumed to have spin-parity and polarization JPM and to
decay via an I-wave into a TITI system of spin S. Therefore the
meaning of the indices is a = JP MIS.
M is the amplitude, symmetric under the exchange of the two nega-

a P
tive pions, for lhe I-wave decay of the J M state into a TI+TI- sys-
tern with spin S. It has been written as
where m 3 rd component of 1
)l 3 rd component of S
8 1 l PI angles and momentum of (TI + TI -2 ) in the 3TI c.m.s.
8 1 l ql angles and momentum of TI +.1n TI.+-
TI2 c.m.s.
AS (sl) is the dip ion propagator, parametrized by TI +TI - reso-

nances of spin S = 0, 1, 2:
180 R. BIZZARRI
with
q 2 'h \
is (S.1') = Vs (-!) ~s
'10 .fS1
and the values of mS and fS being
S = 0 (E) M
0
760 f
0
400
1 (p) M 760 fl 138
2 (f) M 1265 f2 154
We know that the rrrr S-wave interaction is not a simple Breit-

Wigner with these parameters. This however can still be considered
as a reasonable phenomenological representation of the rrrr interac-
tion below 1 GeV.
The data fitted with formula (7) and the values of Pab deduced
as functions of the 3rr mass. The P aa with equal indices give the
intensity of production of state a, the argument of Pab gives the
relative phase between states a and b.
The intensity of the 2+ D wave amplitude and its phase relative

to 1+ waves are shown in Fig. 12. The curves are the expected beha-
viour for the A2 Breit-Wigner and their agreement with the data for
this well-known resonance gives confidence ~n the method.
Fig. 13 shows the dependence from the 3rr mass of the 1+ S

(prr) intensity and of other less important waves. It is seen that
the Al enhancement ~s in fact due to the 1+ S (prr) state. However
the phase of this amplitude vs. other non-resonating amplitudes,
also shown in Fig. 13, is very nearly constant with H3rr contrary
to the behaviour expected for a Breit-Wigner (and in fact observed
for the A2 ). This result therefore strongly suggests that the ob-
served peak is not a resonance.
The method used to describe the 3rr system in terms of the final
state rrrr interaction can be and has been criticized on theoretical
grounds, in fact the amplitudes used violate the requirements of
181
DO MESONS FILL SU(3) NONETS?
0.17<ltI<Q.33 (GeI/C)2
400
i
~300
en
t-
z
~200
100
1.0 1.1
100
Fig. 12 (a) Intensity of the 2T D wave versus 3n mass.

(b) Interference phase of the 2+ D wave with 1+
waves versus 3n mass. Curves are the expecta-
tions from a Breit-Wigner amplitude.
182 R. BIZZARRI
n-p -n-:nt"p combined {2540 Ge\f

GeVk
c
!
200)
fHf~
" S (Qnl
500
ff ~ "0 (QnJ
...
f
~
-
~
U")
N !f 0
! ! 2-P
en OXI
t
~
zu.J (pnJ
f wiltffl~
:>
u.J
:>
II>
,! ~
.++'+
! !
,. P (en)
vsl"P (Qnl
!t
S
u.J
~ ~~{I}iiOQo?Oc.:;\#? ,~+!
~
~
0
-soo
141 ~~~~~~li~\sl'p (enl
u.J
u
zu.J
a:
.... -me ~ . 02
u.J
a: Y ,~~oQ 500 o-s (en)
u.J
~
-1 Q~ vsO-S(enJ
~
Fig. 13 Intensities of different partial waves and

interference phases in the Al region.
unitarity. It is however difficult to see how corrections to this

method could affect so much the phase of the amplitude. A criti-
cism of a more experimental nature comes from the fact that the
amplitudes against which the 1+ S (pn) phase is measured are small
and two of them involve an S-wave nn subsystem which is not well
represented by the resonance which has been used.
In spite of these criticisms we should accept the fact that

the observed enhancement cannot be due to a pure resonance and it
is most probably produced by a Deck mechanism. This does not ex-
clude the presence of a genuine Al resonance, buried somewhere in
this background, but ~n this case we have no clear indication as
to its mass or width (except that it is very unlikely to be narrow).
A possibility is to try to detect a non-diffractive production

of the AI. A recent result, which might be very interesting in
this respect comes from the ACNO collaboration l l ). From 14360
events
induced by 4.2 GeV/c K-, with a cut at u<0.7 GeV 2 (backward 3n

production), they obtain the 3n mass spectrum shown in Fig. 14. A
clear peak is present (besides the A2 ) at m = 1041 13 and with
r = 230 50 MeV. Its production cross-section is comparable with
the backward A2 production and it is almost completely due to 1+ S
(TIl). It is certainly premature to conclude that this is evidence
for AI' since there is no phase determination for this amplitude,
but the result is very suggestive of a resonance.
3.2 I =1
The strange mesons are not eigenstates of C nor of the G-parity
operator. We cannot therefore distinguish between the two states
QA and QB belonging to the two nonets with J PC = 1++ and 1+-. If
however there are two nonets there should also exist two physical
mesons, both with JP = 1+, which can eventually be a mixing of the
184 R. BIZZARRI
Fig. 14 3n mass spectrum from the reaction K p + Z-n+n+n-

for u < 0.7 (GeV/c)2. Events with at least one
n+n- mass combination within 0.62 < m + _ < 0.88 GeV
(p band) are hatched. Curve A shows gh~ expected
distribution with no 3n resonances introduced.
Curve B shows the effect of adding Al and A2 .
SU(3) states QA and QB' I shall indicate these physical states as

Ql and Q2' They are expected to decay into Knn (i.e. K* (892)n
or pK) and should be diffractively produced. In fact a broad bump
has been observed S1nce a long time in the Knn mass spectrum
between 1200 and 1400 MeV (i.e. not far above the K*n threshold)
from Kp interactions without charge exchange.
Also in this case, as for the AI' a sizable Deck background,

mainly K*n, is expected to be present and 1S not easy to disentangle
the evidence for the expected resonances. I shall again discuss
only one experimental result, i.e. the extensive new data on
+ + +-
K-P -+ K- n n p
at 13 GeV obtained with the SLAC spectrometer 12 ). The observed

+
K-n+n- mass spectra are shown in Fig. 15 a) and b), the histogram
representing the raw data and the dots the data corrected for the
apparatus acceptance and efficiencies. Fig. l5c) and d) show the
principal total contributions of the different JP states to the mass
spectrum. These are obtained fitting the data with an isobar
model, similar to the one discussed for the AI' Each wave 1S des-
P n P
cribed by the quantum numbers J M ISO (L), J being the Knn
total spin and parity, M the magnetic quantum number, n the ex-
change naturalitv, ISO denotes the isobar (K*, p, E, K) and L the
orbital angular momentum between the isobar and the third particle
(n or K).
The contributions of the most important waves, and their phase

relative to the l+O+K*n are shown in Fig. 16 for the K*n and in
Fig. 17 for the Kp. These data can be interpreted 13 ) assuming
1) A Deck background l+K*n peaking at ~ 1200 MeV with no associa-

ted variation of phase (dotted curve in Fig. 16);
2) A resonance Ql at ~ 1300 MeV, about 200 MeV wide, coupled mos-

+
tly to pn. This explains the peaking of the 1 pn waves and the
associated phase variation (the reference wave l+K*n is expec-
ted to have constant phase if due to Deck background);
186 R. BIZZARRI
K+
:mo a) h}
Total .'~A..
2000

o
~
.:;-
o 4000 ~
2001 g
~.... "'-.l..-....L..-'--'---'---"O "";;;
..:;. 1
c) d}
~Ii ...
o
1+ o0 0C>::tQ,
:
1+ ."'"
o
o 0
0
o o
o
o

01..lL.JL.:....L----"':.-.c~&J
1.0 1.2 1.4 1.6 to 1.2 1.4 1.6

m(Knn ) GeV m(Kn n ) GeV
Fig. 15 Observed and corrected Knn mass spectrum from

reactions Kp ~ Kn+n-p at 13 GeV/c, and prin-
cipal total JP contributions. The right-hand
scale refers to the observed spectrum and the
left-hand scale only to the corrected spectrum.
3) A second resonance Q2 at ~ 1400 MeV, about 160 MeV wide, coupled

mostly to K*n. This explains the second peak observed in the
l+K*n states at this mass (dotted curve ~n Fig. 16) and the
backward motion of the l+KP phase compared to the l+K*n (Fig.
17), since,if the Q2 is not coupled to Kp, the l+Kp phase is
expected to be constant and the phase of the reference wave
should be moving forward.
Furthermore, the data show a peaking at 1400 MeV of the 2+K*n

wave (Fig. l6e) and f)), most easily interpreted as due to the
decay of the well-known K* (1420) resonance, but the relative phase
+
of this wave with respect to the 1 K*n is very nearly constant in
this mass region, thus indicating that the reference wave is also
executing a Breit-Wigner phase variation.
The coherent addition to the Deck background of two 1+ reso-

nances is therefore required to explain the data. The fitted values
of resonances is therefore required to explain the data. The fitted
values of resonance parameters, according to the SLAC group13) are:

K- It It-" 13 GeV
- 1'< 0.3 GeI/
--;:;---- 1500 a)
--"I:';
c.::J
"":;:;-
::I.
-.." E
.."
150
-:;---- C) d)
;::.
~
..,.
........::...
';,.g
"1==
... 360
'"'"g.
"E 180 ~~
..,."'-
0
I I
0
120 e) fl
..-;:;-- rl+KIt
~
--'fc
~
.D
80
::I.
40
-.." E
.."
0
180
...!!:!
-;;;
en
:s'" Of-- . \~
t
- tf_+~
..,.
r r I
- 180
1.0 1.2 \.4 1.6 1.0 1.2 1.4 1.6
m(K ltltl (GeV) m(Kn: It ) (GeV)
Fig. 16 1+ 0+, 1+ 1+ and 2+ 1+ K*TI cross-sectlons

. an d p_ases
h
relative to 1+0+ K*TI. The crosses denote the K ambi-
guous solutions. The solid curves are from the SLAC
fit (see text). Dotted and dashed curves correspond,
to the background and Q2 contributions, respectively.
188 R. BIZZARRI
600 r------~
a) b)
>
'"
t!:)
o "C
-
';:,E
"C
90 1-
01-
- 90 -
x
+
- 100~~'~1~~L-'~~
240 c) d)
...
~
>
'"
t!:)
270
...
......
'"~
0>
180
'"
:s
90
1.0 1.2 1.4 1.6 1.0 1.2 1.4 1.6

m( Kn;n) (GeV) mCKnn) (GeV)
F'~g. 17 1+ 0 + +a~ d 1 + 1 + pK cross-sect~ons

' '
and ph~ses relat~ve
to 1 0 K*n. The crosses denote the K ambiguous
solutions. The solid curves are from the SLAC fit
(see text).
1290 MeV ISO MeV (Kp dominating)

1400 MeV 140 MeV (K*n dominating)
The statistical errors being small, but the systematic uncer-

tainties quite large, particularly on the widths. For instance a
similar analysis by Bowler 14 ) gives:
and
In this mass region Knn enhancement has been, for a long time,
observed in non-diffractive reactions, the first observation in pp
annihilation (known as C-meson) going back to 1964 15 ). I shall
only mention the very recent results from the ACNO collaboration 16 )
from a 133 events/~b bubble chamber exposure. The reactions
- - +
K P -+;::. (KnTI)
have been studied with a cut ~n the K to _ four-momentum trans-

fer squared u < 1.S GeV 2 . The (Knn) mass spectrum (Fig. 18) shows
a very clear enhancement wi th m = 1276 S MeV and r 78 20 MeV,
decaying preferentially into Kp. It is tempting to identify this
peak with the Ql' but the large discrepancy on the value of the
width is unexplained.
3.3 I = 0
The experimental knowledge of these mesons ~s very poor. For

the 1++ nonet the I = 0 members should have G = + and their decay
into the experimentally more accessible two-body channels (like
nn, KK, wn, nn) is forbidden. possible decay modes are on, Aln,
K*K. A good candidate for this nonet is the D(128S) meson with
r 'V 20 MeV which has been seen in the channels nnn (mostly via on,
followed by 0 -+ nn, and KKn (again dominated by the I = 1 KK thresh-
old enhancement attributed to the 0). Its JP are not firmly estab-
lished but should belong to the unnatural spin-parity series, with
1 + be~ng
.
favoured.
190 R. BIZZARRI
> 150
Q)
:2
0
N
-
"'-..
'"
c:
Q) 100
-
>
Q)
c
....
Q)
..c
E 50
::J
;;;::
1.8
Mass ( Kn n) GeV
Fig. 18 Knn mass spectrum from reactions K-p + ~- (Knn)+
with u < 1.5 (GeV/c)2. The full line curve is
the result of the fit to determine the Ql mass
and width.
Another possible candidate for this nonet is the E(1420) meson

with decay modes KKn, K*K (and charge conjugate) and nnn. As for
the spin parity it is very likely 0- or 1+ but there is conflicting
evidence as to which of these two assignments is to be preferred.
Should O~ be the correct assignment it should probably be consi-
dered as belonging to a radially excited 0 nonet and the second
I = 0 member of the 1++ nonet would be missing.
+-
As for 1 nonet the I = 0 members could very likely decay
largely into pn. It ~s not very easy to pick up such states due
to the presence of one nO in the decay products and the possibly
large width and there are in fact no experimental candidates.
4. Conclusions
After so many years of very hard work the experimental situa-

tion for the L = 1 mesons is still far from being clear. The 2++
nonet is the only one well known. A complete 0+ nonet is emerging
from the data but there are still difficulties in extracting good
resonance parameters from experiment. The observation of two reso-
nances in the Q region is a big step in establishing the existence
of the 1+ nonets but the problems of the nature of the Al and of
the unobserved isoscalar particles are still unsolved.
This situation is in some sense surprising when compared with

the present knowledge on the cc bound systems formed with the
charmed quarks. In this case the narrow width of these states and
their consequent relatively large electromagnetic couplings have
allowed to establish in less than three years the existence and
quantum numbers of a large number of states, the cc system being
now much better known than the older ss.
Is there a possibility in the future of a clarification of

this field? Probably no dramatic forward steps can be expected,
but the progress that has been made in recent years ~s likely to
continue. The use of forward spectrometers can continue to be
192 R. BIZZARRI
very fruitful, the more so if complemented by the possibility of nO

detection.
High statistics (> 100 ev/~b) bubble chamber experiments should

provide further, very interesting, information on non-diffractive
production. A very interesting possibility which is likely to be
exploited in the next years is the study of the L = 1 states as
decay products of higher 1 states obtained via e+e- annihilation,
for instance
or
+ -
e e -+ CP' -+ QK
The ma~n advantage would be to have a situation free from any dif-
fractive background. Good 4n magnetic detectors on low energy e+e-
machines would be necessary.
References
1) S.H. Berman, "Elements of SU(3)" in Symmetries in Elementary

Particle Physics, A. zichichi editor, Academic Press, 1965.
2) J.B. Gay et al., Phys. Lett. 63B (1976) 220.
3) S .H. Flatte, Phys. Lett. 63B (1976) 224.
4) G. Gr~yer et al., Nucl. Phys. 75B (1974) 189.
5) B. Hyams et al., Nucl. Phys. 64B (1973) 134.
6) D. Horgan, Phys. Lett. 5lB (1974) 71.
7) H. Cerrada et al., Grupo de Altas Energias, preprint JEN 76/1.
8) P. Estabrooks, Proceedings of the 1977 Boston Heson Conference

(to be published) .
9) P. Estabrooke et al., SLAC-PUB-1886 (1977).
10) Yu. Antipov et al., Nucl. Phys. B63 (1973) 141.
11) Ph. Gavillet et al., CERN/EP/PHYS 77-15.

12) G.W. Brandenburg et al., Phys. Rev. Lett. 36 (1976) 703.
13) R.K. Carnegie et al., Nucl. Phys. B127 (1977) 509.
14) M.G. Bowler, J. of Phys. G3 (1977), 775.
15) R. Armenteros et al., Phys. Lett. B9 (1964), 207.
16) ACNO collaboration, CERN/EP/PHYS 77-28.

DIS C U S S ION
CHAIRMAN: Prof. R. Bizzarri
Scientific Secretary: A. de la Torre
DISCUSSION No.1
HINCHLIFFE:
In the MIT Bag model of Jaffe et al. the 0+ mesons

are interpreted as (qqqq) bound states rather than as
(qq) states. They say that this explains the large width
of these mesons. Can you comment on this?
BIZZARRI:
I don't have much to comment except that I don't

find it unreasonable to have ann S-state with large
width. Perhaps somebody wants to give a theoretical
argument.
PREPARATA:
The large widths of the 0+ states are far from being

unexpected; here centrifugal barriers are not operative
as in the case of e.g. the At. The establishment of the
0+ nonet is a great triumph for the naive quark model.
LIPKIN:
At the Northeastern Meson Spectroscopy Conference,

David Leith interpreted the 0+ scalar nonet as a four
quark state.
195
196 DISCUSSION
ZICHICHI:
Could you recall to us one argument in favor of one

of the two views?
LIPKIN:
No, I can't.
DEO:
In Flatte's analysis, the analytic continuation to

below the threshold does not appear to have been done
correctly, since a logarithmic factor ~ "1M is missing.
Is this likely to affect your results seriously?
BIZZARRI:
This is the most naive analytic continuation but it

seems to work correctly. Let me ask you a question: do
you think that this analytic continuation is not reliable?
What do you propose instead?
PREPARATA:
This analytic continuation is a little bit cavalier.

One should use dispersion relations, but it reproduces
around threshold very closely what happens for the dis-
persive part.
WETZEL:
t1y question also refers to the 0++ nonet. As far as

I remember the existence of an S wave resonance with I=O
situated underneath the f meson, has been inferred from
combined analysis of ~+~- and ~s~ production data. Why
didn't you mention this finding?
MANDELLI:
A phase shift analysis on ~oio system was showing

evidence for a rather narrow (r tV io M4V) state under
th~ Az/fo peak. Two subsequent experj~ents at CERN on the
~oKo system and at Argonne on the ~+~ system do not supp-
ort the previous conclusions and prefer a solution giving
a broader 0+ wave.
DISCUSSION 197
BIZZARRI:
The last results for the so called ~ have moved

its mass to 1.2-1.3 GeV which is a 1t1t s-wave resonance
in the f region. The elasticity to Ie ~ is very small
as measured by CERN-Munich (see fig .).
WETZEL:
The thing which you show there is not a measurement.

It is a mass independent phase shift analysis.
MIETTINEN:
I wish to mention some new and very interesting data

coming from SPEAR. The figure shows the ~~~-mass distri-
bution in the processes a) "'~ cl> + l1t't'r) b) -- .... GL) + t'1l~u.-).
As we see, the two spectra are very different. In part-
icular, the mass spectrum of the process a) has a con-
centration of events near mtl1TIt-)= lGeV. This may well
be due to the production of the S* resonance. I think
that this data provide support to the idea you discussed
in your lecture, namely, that the strange behaviour of
the R"'"1t- phase shifts near 1 GeV is due to the influence
of a resonance (Sn) in the "i. channel below the physical
threshold.
'41 - .". .".-

.. 5 (a)
~0
2 -
(b) '" - W .". 1r
i30
"en
~20
."-
III
>
III
10
O~~----~--~--~--~
500 1000 I~ 2000 2500
m(.".+ ".-) (MeV/CI )
198 DISCUSSION
BRANSON:
Since you haven't listed any problems with the exist-

ing data, I would like to ask what basic physical princi-
ples can we learn by further measurements of meson reson-
ances and by trying to fit them into multiplets, particu-
larly since we have a good deal of evidence in support of
broken SU(3).
BIZZARRI:
As we could see this morning these SU(3) multiplets
are by no means filled up. As an experimentalist I would
like to find these mesons. One relevant question is whe-
ther the measurement of one particle in a multiplet im-
plies that the entire multiplet exists or whether that
particle was just an accident. This point of view has
been advocated taking for example the deuteron which is
considered to be an accident and it is not sure whether
it has an accompanying SU(3) multiplet or not.
MARTIN:
This morning you mentioned that the problem of Chew-
Low extrapolation was delicate. In fact there are two
problems. One is the extrapolation to get the Pion_Pion
cross section. Another one is that of getting phase
shifts from differential cross sections and this last
process leads to non-unique answers as I explained in my
1975 Erice lecture notes. The only thing which restores
uniqueness is analyticity with respect to energy and this
has in fact been used in practice by Alan Martin and
Pennington. Otherwise you may have two sets of phase
shifts reproducing the same data as has happened in the
past to one of the CERN-Munich experiments.
BIZZARRI:
Yes. There are three relevant points: first to ex-
trapolate the cross sections, then to get a set of phase
shifts. There you have ambiguities but even when you
resolve all ambiguities, which you can by imposing anal-
yticity and constraints from other reactions, you have
to read them in terms of poles in the complex plane and
this third step has still a lot of arbitrariness.
DISCUSSION 199
DISCUSSION No.2
JENNI:
I want to comment that new evidence on the Ai being

a resonance has been presented at the Budapest 1977 Con-
ference by a ETH-CERN-Imperial College-Milano collabora-
tion (J.Pernegr et ale paper submitted to the conference).
Partial wave analysis of coherently produced 3~ in
'It A .... f::>.! 7r1fTC shows resonant behaviour over the A1 region.
BIZZARRI:
This must be new data since it was not presented at

Philadelphia. I was not at Budapest and I haven't yet
received the papers. Does anybody have this data here?
(no response from the audience).
BALDINI CELIO:
Let me put a naive question; is the Deck mechanism

not sufficient to explain the Ai bump?
BIZZARRI:
The bump in diffractive 1tp -.... 3x P is dominated by

the Deck mechanism and shows no phase variation. However
there might be a A, component buried in it which one can
observe in non-diffractive production.
PREPARATA:
I don't understand why a kind of Deck effect should

not be present also in non-diffractive processes, to mask
out the Ai resonance.
MIETTINEN:
That is a very relevant question. You can calculate

this diagram and you get a Deck effect which is not as
strong as in diffractive dissociation.
200 DISCUSSION
MIETTINEN:
I have two questions. Firstly when the SLAC data

on ~~~ production are supplemented by some additional
physical assumptions one may carry out a full mixing
analysis of the Ai nonet. The result is that the mass
of the A 1 should be between 1300 and 1400 MeV and the
width should be very large, r A1 ~ 2.00 MeV . It may well
be that this is the real solution to the A1 puzzle. May-
be people have been looking for the Ai in the wrong mass
region. My first question is: Why didn't you mention
this possibility in your talk?
BIZZARRI:
I forgot to comment on the fact that the two strange
members of the 1~~ and 1+- nonets are expected to decay
into Ie. p and "*'71 with equal probabilities. The observa t-
ion of almost pure states like ICp or IC-*TC. is evidence of
mixing. I do not see how from this mixing between two
different nonets you can deduce the properties of the A1
MIETTINEN:
My second question is the following: You advocated

the idea that the new bump found in the 4 GeV/c ~-p re-
action would be the A 1 One should notic~ that this
state, whatever it is, is produced in a NN annihilation
vertex. The Ai has been searched carefully in charge
exchange processes such as -n. ....... Yl. --. (n+-1t-n O ) " , p and the
existing data puts rather stringent limits on this prod-
uction cross-section. How do you explain the absence of
an A1 signal in this process?
BIZZARRI:
I know of the spectrometer study of the reaction
7t' -p -+ ( 7T "'lcn 0) + t'\
~yy
but I do not think they have yet published a conclusive
analysis of the Ai region. Their data will be very in-
teresting. Of course in this reaction there is also the
1=0 contribution to the 3J(. system. I do not think that
the existing evidence against non diffractive Ai prod-
ution is really so significant. The possibility you are
advocating that the peak seen in the 4 GeV/c ~-p experiment
DISCUSSION 201
would be some sort of "baryonium" state, is difficult to

prove or disprove at present.
MIETTINEN:
I am not advocating anything. I simply like to know

if the estimated cross-section of the A, seen in the
4 GeV/c experiment and the absence of an A1 signal in the
charge exchange processes are compatible with each other.
BIZZARRI:
I don't know.
ANON:
In the charge exchange processes you have a Deck

type background which may complicate the analysis.
MIETTINEN:
Yes, you are right in that there exists a Deck-type

background in the processes. It is easy to calculate,
at least approximately, and is rather small. It does
not change the conclusion. Please, read Gordon Kane's
papers. He discusses the evidence for and against a
relatively narrow Ai in the mass region near 1100 MeV in
the processes mentioned and concludes, that there is
strong evidence against the production of such an A1
with a cross-section theorists would feel comfortable
with.
LIPKIN:
My question has been partially answered so I would

like only to comment that there are equal ~'f> and t:.'*Tt
decays predicted by SU(3) for two states in the Q-region
QAand Qa which are in the same octets as the A1 and B.
Phase space will enhance the ~*~ over the ~f. The
observation of a state which goes only to ~r is evidence
for the existence of both states and strong mixing
between them.
THE PROPERTIES OF CHARMONIUM AND CHARM PARTICLES
Herwig Schopper
Deutsches Elektronen-Synchrotron DESY, Hamburg
and
II. Institut fUr Experimentalphysik der Universitat Hamburg
1. INTRODUCTION
An impressive amount of data has been accumulated over the past three
years which support strongly the idea that elementary particles are composed
of four quarks instead of three. All the predictions based on the existence of
the fourth, the charm quark, have in principle been borne out by experiments, and
from detailed investigations very interesting results could be obtained for the
strong and the weak interaction. Most of these results have been obtained with
electron-positron storage rings, and only these will be discussed in this re-
port. Additional data from hadronic or neutrino interactions are in general agree-
ment with the e+e- data and will not be discussed here. This series of lectures
will be entirely devoted to a discussion of mesons, i.e. quark - anti quark
systems. Although some indications have been found for the existence of charmed
baryons, these data are still rather scanty. In the last chapter we shall also
discuss the upsilon particle.
203
204 H.SCHOPPER
1.1 Experimental techniques

1~!! __ ~!~~!~2~:~2~!!~2~_~!2~~9~_~!~g~
Most of the data have been obtained at SPEAR in Stanford, California, and
DORIS at DESY, Hamburg. SPEAR is a single ring machine in which electrons and
positrons circulate in the same vacuum chamber in opposite directions. The stored
particles are concentrated in one bunch per beam, thus colliding only in the two
intersection regions. DORIS,on the other hand, consists of two rings, one on top
of the other. The beams cross at an angle of 24 mrad. As a consequence,
each beam can contain up to 480 bunches, yielding higher luminosities at lower
energies,where the luminosity is limited by space charge effects. At higher ener-
gies,where the rf power is the limitation, the single ring operation is more
favourable.
The two rings of DORIS would make it possible also to study electron-electron
and electron-proton collisions. These options have not been used so far, because
of the strong interest in e+e- physics.
The layout of DORIS and the injection scheme shown in fig. 1.1. So far DORIS
mostly has been operated at energies between 2 x 1.5 and 2 x 3 GeV. The electron and
positron currents are usually about 200 mA and the beam lifetime is varying
between 5 and 12 hours. Currents between 0.5 and 0.8 A have been achieved, but
the beam lifetime then becomes short and the momentum smearing in the beam gets
bigger than 1 MeV. In October 1977 DORIS was converted to single ring opera-
tion and energies of 2 x 4 GeV have been achieved. The limi-
tation is then given by the rf-power. It is planned to add two more PETRA
cavities in DORIS, which should make it possible to push energy up to
2 x 5 GeV. This will be tried during 1978.
PROPERTIES OF CHARMONIUM AND CHARM PARTICLES 205
r:::. ~- _ _
~ Electrons
~ SYNCHROTRON
m~~'\~ , Positrons
W :Interaction Point
LlNEAR-~ \\
ACCELERATOR I
r' '
LlNEAR-
ACCELERATOR 2
600 MeV
Fig . 1.1: layout of DORIS and injection scheme
In fig.1.2 luminosities as obtained with

DORIS are shown. Going up in energy, the
number of bunches is successively reduced,
giving a tooth-structure. The dashed line
indicates the luminosity expected for
single ring operation. The crossover
of the two curves is around 2.5 GeV per
I.
I I
480burllChn ?-,O bunc h"
beam, indicating that at lower energies
multi-bunch operation is more favourable
, ...r "'7L~liEl(I'~O."'1
than single ring operation. L~E'
.7
, E(GeV Ibe.3m
206 H. SCHOPPER
[nb'~ ,
AVERAGE WMiNOSf1YIWEEK
OF OORIS
f
\00 8 t---
I~' ~
1~0
~
l- e- t1 t---
o JJ1l
r
I() liJ 30 III
1976 --~ 19n--
\0
j
I() WEEK
Fig. 1.3
Fig.I.3 shows the average luminosities per week obtained during '76 and beginning
of '77. It is the average luminosity which is really important for the experi-
ments. The continuous increase of the average luminosity is mainly due to im-
provements in the stability and reliability of the machine. With a total cross
sect i on of about 40 nb several hundred events can be observed per day.
l~!f_~g~~trQ~~t~r~
In order to observe and analyze the particles produced in e+e- annihila-

tion magnetic spectrometers are used in most experiments.
ng . I.4 shows a blow-up diagram of the magnetic detector at SPEAR. A coil

of about 3 m diameter produces a longitudinal field parallel to the incident
beam, Cylindrical chambers permit the detection of charged particles, providing
MUON TOWER
r/J -3m
t -3m
, kr
SPEAR SLAC-LBL MAGNETIC DETEClOR
eo'eo- COLLISIONS
Fig, 1.4
Iron yo~.
"==01 -----tlt+--+-..,J,~
JO~lI'< 1'11<,"
l.I05crn
TtllXlOf Ace:
6,R .81 ,X"
"" ',K ~ 2.0 II.VI<
Fig. 1.5
very good momentum analysis, which is one of the strengths of this spectrometer.
A solid angle of about 65 % of 4n is covered. Electrons and y-rays can be ob-
served by shower counters, which are placed outside the coil. Time-of-f1ight
measurements with scintillation counters make it possible to identify particles,
and in particular the separation of K and n'S turned out to be very important
for the discovery of charmed particles. Large spark chambers located outside
the iron yoke allow the detection of muons. Since the iron yoke is not very
thick, there is however some punch through from hadrons. In order to provide
cleaner muon detection for at least a limited solid angle, additional concrete
absorbers were later added on top of the detector ("muon tower") . Very
recently a lead glass wall with an active converter has been added, which has
proved to be very important for the clean detection of electrons.
Fig.1.5 shows the PLUTO spectrometer at DORIS. PLUTO also uses a longitudinal
magnetic fie1d,which, however, is produced by a superconducting coil giving a
field of about 2 tes1a. Since the diameter of the coil is only about 1 m, the
momentum resolution is lower than that of SPEAR and time-of-flight measurements
cannot be used to identify particles. On the other hand the cylindrical chambers
inside the coil cover about 86 % of 4n. The cylindrical chambers are interspersed
with two lead converters (0.4 and 1.7 radiation le~gths thick, respectively),
which allow the detection of y'S and the identification of electrons. Large
chambers outside the iron yoke are used to detect muons. Since the iron yoke
208 H. SCHOPPER
provides a hadron absorber more than 60 cm thick in all directions, very good
muon identification can be obtained. Indeed, the misidentification of a hadron
for a muon is less than about 3 %. Also electrons can be identified quite re-
liably, again with a misinterpretation probability of only a few percent. These
properties have been very important in verifying the existence of a heavy
1epton.
Fig.l.6 shows a schematic view of the double arm spectrometer DASP at

DORIS. Here two big magnets provide a transverse field outside the interaction
region. Particles entering the gap of the two magnets can be analyzed very pre-
cisely (momentum resolution better than 2 %). Since long distances are involved,
particle identification by time of flight is possible. Electrons and yare again
identified with shower counters and v are observed behind iron absorbers. Thus
very good identification and momentum analysis of hadrons is achieved
over a solid angle of 0.9 sr. Two cerenkov counters have been installed covering
the magnet gaps, which allows a very clean electron identification. A central de-
tector consisting of proportional tubes and shower counters but without magnetic
IRON ABSORBER
~IRON
Q
ABSORBER SHOWER
COUNTER
Fig. 1.6. Spectrometer DASP

analysis covers about 70 % of 4~. This central detector allows the determination
of the direction of y's and charged particles with an.accuracy of about 20 , which
has been very essential in some of the experiments.
Several non-magnetic detectors using NaI cyrstals, lead glass counters, or

large neutron counters have been used at SPEAR as well as at DORIS. Unfortuna-
tely, there is not enough room here to describe all the experiments.
!~!~ __ I~Q~~_Qf_~~Q~r!~~~~~
If an electron and a positron annihilate, a virtual photon is produced
which then decays either again into a lepton pair or into hadrons (fig. 1.7).
The total final state must have the quan-
tum numbers of the photon JP = 1-. Most
a) eTtct
inclusive spectra of the experiments which have been per-
formed and which will be described later
fall into the following three classes:
1. Non-resonant hadron production (fig. 1.7a).
Here one might ask questions like what is
the multiplicity of produced hadrons, how
does the total cross section as a func-
tion of energy (excluding resonances) vary,
what are the inclusive spectra of diffe-
rent kinds of particles like, etc.
2. If the energy of the virtual photon
coincides with the rest energy of a
vector particle having the quantum num-
bers of the photon, the hadron production
Fig. 1.7 will be enhanced. Measuring the total
cross section or special channels one will notice a resonance-like structure
if the data are plotted as a function of the e+e-energy (fig. 1.7b). This is a
very powerful method to detect new particles; however, only particles with
the quantum numbers of the photon can be discovered in this way.
3. Other states can be found in the following way: assume that the virtual
photon decays into two resonances (fig. 1.7c) which in turn decay into
other hadrons. If the momenta of the final particles are measured, then
one can combine two or more particles and calculate from the relativistic kinema-
210 H.SCHOPPER
tics the invariant mass of the object from whose decay they originate. Since
one does not know a priori which particles come from the decay of the resonance,
many wrong combinations of course are taken, which produce a smooth background.
The right combinations of particles,on the other hand, produce a sharp peak on
top of this background. In this way states with other quantum numbers can be
found; of course it is necessary to measure ~ decay particles in order to
determine the invariant mass.
4. The kinematics allow us to calculate also the mass of the system recoiling
against the invariant mass determined as described. If the recoiling state also
contains well defined resonances, peaks in the distribution of the recoil mass
will be found. Since the recoil mass can be calculated from the momentum of
the first resonance and the initial state, it is not necessary to observe the de-
cay products of the recoiling resonance.
1.2 Phenomenological Models
The purpose of this review is to summarize the experimental facts and not
to discuss various theories. To interpret the data, phenomenological models
will be used which have been developed over the past yeal's and which are based
on simple concepts. Most of these models lack a rigorous foundation, and
hence they are criticized by many theorists. On the other hand, these models
have exhibited a surprising and extremely successful predictive power, and there-
fore the usefulness of such models is beyond any doubt. Before discussing these
models, it might be useful to recall some generally accepted ideas.
1~f!_Q~2r~_~Q~~1
The following discussions will be based on a SU(4) xSU(3) quark model im-
plying that quarks have 4 flavours (u,d,s,c) and 3 colours (blue, red, yellow).
One of the main questions will be to clarify if the experimental results are in
agreement with the existence of a charm quark, which has been requested to
restore quark-lepton symmetrylo6) and to explain the absence of neutral currents
in K decay107).
The colour SU(3) is supposed to be a perfect local symmetry, implying that

only colour neutral (white) particles can be observed. The following arguments
support the existence of a colour degree of freedom:
1) to explain the spin-statistics of baryons consisting of 3 quarks,

2) to obtain a non-Abelian group,
3) to explain some experimental results, e.g. decay probability of
nO~ YY (factor 32) and the
ratio R = (e+e- ~ hadrons) I (e +e- ~ w+w- )(factor 3),
4) to understand why particles consist only of ql q2 (mesons)
or ql q2q3 (baryons).
The quantum numbers of the quarks are
flavour Q 13 S C Y
u 2/3 1/2 0 0 1/3

d - 1/3 - 1/2 0 0 1/3
s - 1/3 0 - 1 0 - 2/3
c 2/3 0 0 1 - 2/3
and baryon number B = 1/3, spin = 1/2 for all quarks with Q = 13 + i (B+S+C).
The SU(4) flavour symmetry is broken by the different quark masses.
Since experimental results on charmed baryons are still rather scarce,

only' mesons will be discussed. They consist of a quark and an antiquark. This
leads to the group theoretical reduction
8 +1+3+ 3+ 1
with C 0, 0, -1, +1, O.
The octet and singlet with charm charge C = 0 correspond to the old SU(3)
symmetry. The two triplets require the existence of 3 particle states with
charm C = +1 and -1, respectively. Finally the last singlet is associated
with a state ce, the two charm charges cancel and hence C = (1 ("hidden charm").
!;~~_~~rQ~Qg~~~~!~~_g~~
QED can be generalized to a gauge field theory with local SU(3) colour
symmetry 1). The interaction betvieen quarks is mediated by gluons in analogy to
photons in the electromagnetic case.
212 H. SCHOPPER
~
{ -------g
e vadfilIT'
e
>------y
UCD o..ED
The main differences are:

a} there are 8 massless neutral gluons with a quantum number colour
(compared to only one neutral, colourless photon);
b} the coupling constant as is a function of the interaction energy E
(instead of being a universal constant).
On the basis of very general assumptions one obtains 2}
as(EO}
a (E) =--~25--~(-E-}--ln--(-E/-E~}2~ ( 1.1)
s 1 + T2iT as 0 0
where E is an arbitrarily chosen normalization energy. The factor 25/12n is

o
derived from dimensional counting*) and this particular value is associated to
4 quarks. The ln term corresponds to vacuum polarization in QED.
Equation (1.1) implies that as becomes small at high energies (corresponding

to small distances) and gets big at low energies (large distances). This has
important consequences:
a} Asymptotic freedom: If the distance
between two interacting quarks gets very
small (the interaction energy is high)
the coupling constant as < 1 and one
L-.-4-----E has a similar situation as in QED. One
Eo
gluon exchange prevails and perturbation
theory can be used. ~1any of the formulae developed for the hydrogen atom or
positronium are applicable.
*} 33 - 2N, where N is the number of fl a vours.

b) Quark confinement (infrared slavery):

At large distances the interaction energy is small and according to (1.1)
as gets large. The strong coupling might prevent the liberation of quarks
since the production of qq pairs is more likely. However, it is not yet well
understood how this confinement mechanism works and in particular it is
not clear whether it can be understood in terms of many gluon exchanges.
In the framework of a locally broken but globally conserved SU(3) colour
symmetry free quarks and gluons might exist with quite strange proper-
ties 3a ) .
1.23 Potential models
For the interpretation of the experimental results various potential models

have been developed. The choice of the q-q potential is based on the general ar-
guments given above. In particular the Hamiltonian constists of 3 parts 4)
(1. 2)
+ L( r)
The first 1ine is, of course, the free particle term, where ml and PI stand for
the masses and momenta of the 2 quarks. The second line is associated to the
short range interaction whose radial and spin dependence expressed by S12 are
assumed to be the same for the Coulomb force and the strong interaction. The
only difference is the replacement of the fine structure constant a by as of
(1.1) and of the electric charges Ql Q2 by a colour factor k which is
k = - 4/3 for a qq system and k = - 2/3 for a three quark state.
The third line of (1.2) stands for the long range binding potential. A
preferred guess for L(r) i s 5)
L(r) =a r (1. 3)
i.e. a linear dependence on r and the constant a has to be determined from

experiments. Sometimes 6 ) a harmonic oscillator potential L(r) = a r2 is used in-
stead of (1.3). Practically nothing is known from a theoretical point of view
on the spin dependence of L(r) and on the influence of the quark masses. Also
214 H. SCHOPPER
the properties under Lorentz transformations are under question. A first guess
might be that L(r) transforms like a Lorentz vector but scalar contributions are
possible. Various models differ by the assumptions made for L(r). Recently also
logarithmic potentials have been considered l34 )135).
The simplest model is based on the following assumptions:

a) non-relativistic potential (relativistic corrections up to
(v/c)2),
b) S12 is Coulomb-like, i.e. l/r-behaviour, spin-spin, LS and
tensor couplings,
c) L(r) = ar, Lorentz vector.
The explicit form of S12 is 3 ,4)

.... ....
(1.4 )
1 1........
- - {-2 (r x PI)
2r 3 m1
The second term in the first line describes the spin-spin interaction
giving rise to the hyperfine splitting, the third term gives the tensor
and the second 1i ne the Lx S coupl i ng. The thi rd 1i ne ori gi nates from re:-
lativistic corrections of order (v/c)2. The last line has no classical
analog, its origin is the reduction of the relativistic ai-matrices to
the Pauli 0i spinors.
!;g~g __ ~2r~_~2E~i~~i~~~~_~2~~1~
a) General potential V(r)
Some authors 7 ,8)consider a more general r-dependence of the potential
V(r) than the Coulomb like l/r. In this case the gluon propagator has
to be modified
)J 1
Yl)J Y2 . 17 -- Yl)J Y2 )J v(k2)
where v(k 2) is the Fourier transform of V(r).
For simplicity S12 is given only for the case of equal masses of the two
interacting quarks:
4 = _1_ (d 2V
- "3 Ct s S123m2
_ 1.dV ) [->
-> -
s 1 ->s2 - 3 (s1 r ) (~2;) ]
dr r dr
2 ->->
+- (s1 s 2) ,,2 V( r) (1. 5)
3m 2
-> ->
+ _3_ (1 dV (
2m2
r dr r x -p (s1 + ~)
+ spin independent terms

Ct
if Vir) = - rS + ar: also confinement potential gives spin dependent effects!
The spin independent terms are the same as in (1.4), last two lines.
For V(r) = - Cts/r equ. (1.5) is transformed into (1.4). If the "naive"
potential V(r) = - Cts'r + ar is chosen then the confining potential contributes
automatically to spin dependent effects.
The level ordering for quite general classes of potentials has been studied
by Grosse and ~'artin8b).
b) Lorentz properties of L(r)

Since very little is known about the long range potential L(r) it cannot
be taken for granted that it behaves like a Lorentz vector. A more general an-
satz is a mixture between vector and scalar 9,10)
( 1.6)
where t. are unit matrices.

1
For a linear potential
L(r) = (Kv . Yl ll Y2 1l + Ks .1 1 1 2 ) r (1. 7)
where the constants Kv and Ks determine the mixing ratio.
c) Anomalous Pauli coupling

In analogy to the anomalous magnetic moment it might be considered that
quarks possess anomalous gluon couplings 11 ). This implies that one makes the
following substitution
216 H. SCHOPPER
where aeis the anomalous moment. As a result expression (1.5) has to be modi-
fied in the following way:
4 S-~ + (tensor term)

- 3 as 12 - 3m 2
2( 1+~2
+ 3m + (spin-spin term) ( 1.8)
3 24':.
+ zmz (1 +~) + (LS-term)
+ unchanged term.
From these coefficients it follows immediately that an anomalous coupling qe has

a larger influence on the tensor and spin-spin terms than on the LS-coupling.
Some people question why a fundamental structureless particle like a quark

should have anomalous couplings. It does not seem clear if they could arise from
higher order gluon terms 12 ). However, some confinement mechanisms seem to require
~F O. For example the MIT bag 13 ) requires (1+~ ~ 0 if the quark mass m ~ O.
d) Coupled decay channels

For a bound system like charmonium the decay channels above the binding
energy (e.g. cc ~ cu + uc) are neglected in the first approximation. However,
just below the dissociation threshold the virtual channels can modify the
bound state. Such corrections have been calculated5 ,14).
e) Annihilation graphs
In the charmonium system virtual transition cc ~ y ~ qi qi are possible,
where qi are other than c-quarks. These virtual transitions effect the bound
cc state and lead to non-negligible corrections 14 ).
The influence of the improvements of the standard model will be discussed
below for particular measurable quantities like level splittings, transition
probabilities, etc.
!~~~ ____ rQ~Q_!b~Qr~!i~~l_~Q2~l~

Since the potential models are not relativistically invariant, even
if relativistic corrections are included, some authors derive level schemes
from group theoretical models. The group 0(4) for example corresponds to the
relativistic dynamics of charmonium 15 ). Also broken SU(4) has been in-
vestigated 16 ) .
These models yield level schemes with quantum numbers which differ from
the standard model. However, it seems that they did not really help to solve
some of the difficulties which will be discussed below and hence these models
will not be discussed further.
218 H. SCHOPPER
2. MASS SPECTRA
Experimentally the masses of the new particles are determined by three

classes of measurements (chap. 1.13):
1. Peaks in the cross section for e+e- -> hadrons as function of the c.m.
energy (fig. 8b). Since in this case the virtual photon is con-
verted into the resonance, its quantum numbers must be the same, i.e.
only JP = 1- particles can be detected in this way. As is well known,
the J/~, ~' and the masses of higher excited states were determined
in this way. For bound states the resonances are very narrow (narrower
than the experimental resolution of a few ~leV), whereas above the DO
production threshold the resonances are several hundred MeV wide.
2. The masses of particles with other quantum numbers can be found as
invariant masses calculated from the momenta of their (charged) decay
products (fi g. 1. 7c) .
If the resonance decays into the particles 1 and 2 whose momenta en-
close the angle 8 one finds the invariant mass of the resonance from
( 2.1)
Since a priori it is not known which particles in the final state ori-
ginate from one resonance, one has to try all possible combinations.
The wrong combinations and pure phase space decays produce a slowly
varying background in the distribution of the invariant mass. Ex-
perimentally mass resolutions of the order of 20 MeV can be obtained.
Sometimes the identification of the particles in the final state is
not or only partially possible (e.g. ~-K separation). In such a
case the wrong assignment of a particle mass to a certain track
leads to "kinematic reflections" in the distribution of invariant
masses.
A few special cases are of particular interest. If the decaying system

with mass Mo is at rest PI -P2 P
= = hence the Q-value of the decay is given
by
Q= ~10 - m1 - m - p2 { 1 +1 } (2.1a)
2 - ,z
m1 + vp + m12' m2 + y['pc + m2C'
For p m 1 , m2 the Q value is proportional to p2 and hence a comparatively

crude measurement of the momentum yields quite accurate values of Mo' This
procedure is useful if particles are produced just above threshold, e.g.
e+e- ~ ~(3.77) ~ DO.
-
If a particle moving with momentum p decays into 2 photons the most

likely decay is the symmetrical one where the 2 photons have the same energy
k and the angle between each photon and p is the same I7 ). For this case one
finds
cos8 /2 = p/2k = (p/2) . jm 2 + p2' (2.2)
where p and m are the momentum and mass of the decaying particle. The symmetri-
cal decay angle 8 is also the minimum angle associated to a particular m and p.
A cut-off in the angle can therefore help to distinguish between different
particles.
A special case are decays with 3 photons in the final state, e.g.
J/~ ~ yX ~ y(X ~y). Combining (2.1) and(2.2) one obtains for the invariant
mass of the X-state
Mx = E sin 2(8/2) / (1 + cos8/2), (2.3)

0
where Eo is the total energy.
This implies that the mass can be aetermined from a measurement of the direction
of the photons alone and this is still true in the general . case of an asymmetric
X-decay.
3. If not all of the decay products of a resonance can be detected (either be-
cause they are neutral or do not fall into the acceptance of the spectro-
meter), it is still possible to determine its mass if the resonance in question
R2 (fi.1.7c)is produced together with one other resonance Rl or particle.
From the masses mi and momenta Pi of the decay products of Rl the recoiling
mass can be calculated.
220 H. SCHOPPER
(2.4)
If the recoiling mass is associated to a resonance, one finds a peak in the

recoil mass distribution.
The results of such experiments are summarized in Table I. The way how masses
and the quantum numbers were determined will be discussed below when the produc-
tion and decay mechanisms of the new particles is described.
Table la: Masses of charmonium states
Name State Mass (MeV) Name State ~1ass (MeV)

J/1jJ 1 jS1 3096 2 X =l1c ? 1 is ? 2830 30

lji' 2 \ 1 3684 5 l1'c ? 2 Is ? 3454 7
0
X
1 3p 3414 4

Pc 1 3p 3508 4
1
1 3p
____ 2_ 3552 6
- - - - - - - - - - - -- - - - - - - -
X
------ ---------
lji' , 3 3S ? 4028 1 30 3.772 3
1 1
- 4150
4414 5
Table Ib: Masses of charmed particles
Name State ~1ass (MeV)
0 IS 1863.3 0.9
0
0+ IS 1868.3 0.9 MARK 1104 )
0* 3S 2006 1.5
1
OH 3s 2008.6
1 1.0
F+ IS 2030 60
OASp105)
FH 3S 2140 60
1
2.1 Masses of JP = 1- c~ states
We now want to compare the experimental mass spectrum of the cc- system
with the theoretical expectations based on the simple models described in
chapter 1.2. For the Hamiltonian (1.2) one expects a hydrogen (or positronium)
-like level scheme, as shown in fig. 2.1. One has different ladders for diffe-
rent angular momenta ~ and the levels are split due to hyperfine splitting,
spin orbit and tensor couplings. For a pure Coulomb potential the 2s and 1p
states would be degenerate. Because of the confining term (1.3) in (1.2) the p
and d-states are shifted to lower energies. Indeed for a r 2_potential (harmonic
oscillator) the 1p-state would lie in the middle between the 1s and 2s-states.
The lowest d-state has JF = 1- and hence could interfere with the 2s 1-state,
if the shift is large enough.
Ortho tt Para tt
35 J=r (1jJ') - - - ----
~ )=2+
J--O- 2p
$ =1
~ _
---'----<..s-o-\.:J:. $=1 1=3
J:1+ ~
s:-o-YI
J=2" -
1
~
$=J=r(~) t HFS
$ =J =0-
l= 0 2
Fig. 2.1: Level scheme for charmonium
222 H. SCHOPPER
Soon after the discovery of the J/~ and ~' particles several authors
solved the non-relativistic Schrodinger equation with the potential
as r 2
V(r) = - -
r
{1 - (-)
ro
} (2.5)
Identifying J/~ with the 13S1 and ~' with 23S1 state the following parameters
were found 5)
a s (3.1 GeV) = 0.2, ro = 0.2f, mc = 1.6 GeV/c 2
These values justify qualitatively the assumptions on which the naive model
was based: the Coulomb-like part is indeed short range, the coupling constant
is smaller than 1 and the c-quark mass is large (non-relativistic Schrodinger
equation). On the basis of these parameters the masses of higher lying non-bound
resonances were predicted 14 ):
state prediction expo (Table 1a)

1201 3.75 3.77
33 S 4.2 4.15
1
33 0 4.6 4.41
1
Keeping in mind the simplicity of the model, these predictions are extraordinarily
impressive. The difference between prediction and experiment for the two highest
states can be attributed to neglecting coupled channels 14 ) (see 1.232).
2.2 HFS-Splitting of S-levels
Because of the spin-spin coupling the states with opposite and parallel spins
are split. If the X-particle found by OASP is identified with the 11so state (usu-
ally called nc) and the resonance at 3.45 GeV with the 21so state (n c ') (table 1a)
one finds the splittings
6 (~' - n'c) = 230 MeV.
These large splittings have been considered for quite some time as a major diffi-
culty for the charm model, since most calculations produced much smaller
spl ittings.
If the spin-spin coupling is associated only with the short range part of the
potential, one finds
a) for Coulomb-potential 7,8) V(r) - j a/r

32'11 a
6Mshort = _ _s 11jJ(0) 12 (2.6)
HFS m1 m2
where m1,m2 are the masses of the bound quarks and 1jJ(0) is the wave
function at the origin.
b) for general short range potential V(r)7,8)
short _ 2 2
6MHFS - 3 m1m2- < V V( r) > (2.7)
c) If spin-spin coupling is assumed to exist also for the long range part
of the potential, one obtains 18 ) for V(r) = ar
6M lon g = i (1 +de1) (1 +2) a <l> (2.8)
HFS 3 m1m2 r
Here anomalous couplings~, ~2 (see 1.232) for the long range force have been
included. The total splitting is
6M = 6Mshort + 6M lon g (2.9)

HFS HFS HFS
g~g!_~Eli~~i~g_2f_sb2r~2~i~~_~~2~~~
The experimental ratio RHFS = 6(1jJ-n c ) /6(1jJ' - nc ') = 1.2. According to (2.6)
one has RHFS=I1jJ(O)I~/I1jJ(O)I~,. The functions at the origin of the J/1jJ and 1jJ'
particles can be inferred from their leptonic decay widths and one finds
RHFS =r1jJ(e+e- )/r1jJ,)e+e- )=3.9/2.4=1.6. The agreement is not so bad and supports
the simple model.
For the absolute values of AM HFS on the other hand one calculates much
too small numbers on the basis of the standard model. Splittings of about 30
to 80 MeV were obtained 2,3,4,7) depending on various assumptions and on diffe-
rent values of ~(O). Using a more general potential V(r) the splitting could be
increased 8 ) to A(~ - nc) = 120 MeV and A(~' - n~) = 92 MeV, still too small by a
factor of 2. With special potentials the right values could be derived but this
leads to unnatural conditions for the lept0nic decays19).
The easiest way to explain the experimentally found HFS-splitting seems

to be anomalous gluon coupling analogous to the anomalous magnetic moments.
224 H. SCHOPPER
Schnitzer 18 ) assumed that this coupling is small for the light quarks (~,
~d ~ 0) but appreciable for the charm quark (at c : 1). With such a large ~
the HFS-splittings and also the LS-splitting of the P-levels come out some-
what too bi9 11 ) *). A value Zc '" 0.4, i.e. (1 + rc)2 2 in equ. (2.8), reproduces
the experimental results better.
Recently it has been shown 62 ) that instantons may generate a spin-spin

interaction between quarks. Here small quantum fluctuations about the per-
turbation-theoretic vacuum are replaced by a coherent superposition of vacua
with different topological character. A quantitative estimate of these effects
shows that the spl itting between J/lj! and nc may be dominated' by them and this
splitting might be a direct evidence for the existence of instantons.
g~gg_~~!i!!i~g_Q!_g_~~~_~_~~~Q~~
It is very interesting to apply the idea of anomalous moments to the 0
and F mesons. From (2.5), (2.7) and (2.8) one derives
,',(0*-0) (10030) MeV + (1+~) (1+~) 114 MeV

(2.10)
,',(F*-F) (65 20) MeV + (1+~) (1+~) 144 ~1eV.
Since in the meantime the experimental HFS-splittings of the charm mesons

became known (table 1b) one can derive information on the ~i from (2.10). With
~(O*-O) = 145.3 MeV and ~(F*-F) = (12040) MeV one finds (1~u)/(1+as) ~ 1 and
1+~ :::1.4 as needed to explain the charmonium splittings (2.10) yields
(1+tU) ",(1+~)$0.25. As a consequence the anomalous coupling is small for particle~
consisting of light quarks since (1+,~J ~ (1+,j~;l2 < 0.06; it contributes somewhat
for the charmed mesons 0 and F since the mixed terms are of order (1~u) (1~)
~ (1+.ts) (1+JC) '" 0.3 and it is most important for charmonium because (1+,Fc) 2 '" 2.
It seems surprising that 1+~ and 1+d$ should be so small. However, this fits very
nicely with some ideas about quark confinement. For the MIT bag it has been shown 13 )
that (1+~ + 0 if m + 0 and hence it seems plausible that 1+~ is small for the
light quarks. Of course it would be nice if the anomalous gluon coupling could
be derived from higher order terms in analogy to g-2 of the electron. The lowest
*) ~(~-nc) = 300 MeV, ~(~'-nc) = 250 MeV,

~(3p2-3p1) = 125 MeV, ~(3p1-3Po) = 141 MeV.
terms give 12 ) l+Z"'l + (4/3) (Ct/211) = 1.04, which is much too low. The matter
gets quite complicated if the long range behaviour is included. In that case
divergencies appear and arbitrary cut-offs have to be made.
Fritzsch 20 ) also considered the possible existence of anomalous gluon

coupling and taking the analogy serious between colour and electromagnetic mo-
ments he derived:
I'I(F* - F) jJs - 3 jJ (1\.)

(2.11)
1'1(0* - 0) jJu jJ (p)
where the jJ are the total magnetic moments. With jJ(p) = 2.79 and jJ(l\.) = -0.670.06
one expects for the ratio (2.10) the value 0.72 0.06 which is in excellent agree-
ment with the experimental ratio 120/145 = 0.82 0.3.
Finally it should be remarked that the introduction of an anomalous gluon

coupling does not change the electromagnetic transition rates since the gluons
carry no electric charge.
The splitting between the charged 0+ and the neutral 00 meson has been esti-
mated by various authors. If the isospin breaking is calculated with a non-rela-
tivistic mode1 23 ) one finds 0+_0 0 '" 15 MeV, whereas a more refined mode1 24 ) yields
a value of 6.5 MeV in excellent agreement with the experimental value 5.1 2.8
(see table 1b).
2.3 Splitting of P-states
Let us now turn to the splitting of the triplet P states which is associated
to LS and tensor couplings (see 1.23). For these states the quark spins are parallel
and therefore the total spin S = 1 couples with the orbital angular momentum ~ = 1
to total spins J = 0, 1, 2. From equation (1.5) one can derive the following ex-
pressions for the masses of the triplet P-states:
226 H. SCHOPPER
3 2
P2 =A + B - "5" C
3 (2.12)
PI = A - B + 2 C
3p = A - 2B - 4 C
o
with B = 3 < 1 dV > LS-term

W r Or
_ 1 < 1 dV _ d2 V > tensor-term
C- T2iijT r dr d?
A arises from the SS-term and the spin independent terms. It is useful to
define the ratio
which assumes the following values:

R
..-L
short range Coulomb potential (like positronium) 4/5
standard potential (2.4) 1.2
linear potential 1.4
harmonic oscillator potential (C = 0) 2
For the potential (1.5) the possible range is 0.8 ~ Rp ~ 1.4.

From table 1a one deduces from the experimental masses
3P2 - 3P1 = 44 MeV

Rp = 0.47.
3P1 - 3Po = 94 MeV
This presents a serious difficulty since this value is outside the theoretically
acceptable range. From the experimental splittings and (2.12) one finds C/B ~ 0.3
implying that tensor forces have to be taken into account.
Difficulties arise not only for the ratio Rp but also for the absolute values
of the splittings. Associating the spin effects only with the short range Coulomb
potential gives splittings which are too small by more than a factor of 53). Includ-
ing the long range potential one obtains 35 l4 ) about the right value for the 3P2- 3P l
difference but the ratio Rp comes out wrong. The coupling to decay channels 14 ) and
anomalous moments have little effect (see 1.232) and cannot explain the big dis-
crepancy in Rp'
The only way proposed so far to remedy this difficulty is the assumption that
the long range potential L(r) is not a Lorentz-vector but contains scalar contri-
butions 21 ) (see 1.232,equ. 1.7). In this case the ratio Rp can have any value. Indee
for a pure scalar L(r), i.e. Kv = 0 in equ. (1.7), the ordering of the 3p states is
reversed, with the 3Po being the highest state, unless unreasonable values of as
are permitted 22 ). The experimental ratio Rp ~ 0.5 is reproduceo21 ) with
Ksl (Ks + Kv) ~ 0.8 implying that the long range potential is mainly scalar. It
should be noted, however, that Ks F 0 reduces the HFS splittings.
A few additional remarks will close the subject of mass splittings.

The singlet P-state lPl is expected to coincide with the center of gravity of the
3p states. The 1P1 state has not been seen ye~ being a 1+ state it cannot
be produced directly in the e+e- annihilation and is also hardly accessible by de-
cays from higher states.
The 13D1 state is expected to be suppressed by spin effects and coming close
to the 23S1 these two states are likely to interfere. This will be discussed further
in chapter 4.1.
2.4 Summary of mass spectra

a) Experimentally almost all of the levels predicted by the charm model have
been seen. The 1P1 state is still missing but this fact is easily to be
explained.
The existence of the X particle at 2.83 GeV is now well established ana
its mass value poses no serious problem. Its identification with the
11so state,however, creates some troubles as far as decay rates are con-
cerned (see 3.1).
The nature and existence of the state at 3.45 GeV has to be verified
and its interpretation as 21so is in question.
b) The potential Vir) = as/r + ar describes quite well the position of the
bound charmonium states as well as the charmed particles. Indeed, some
spectacular predictions could be made. Corrections due to coupled decay
channels and virtual transitions are not negligible.
228 H. SCHOPPER
c) The HFS-splitting for the charmonium states and the D and F mesons can
be understood in a common way if anomalous gluon ("magnetic") moments
ilre introduced for the long range force. Alternatively this splittino
mi9ht be explained in terms of instantons.
d) The experimental splitting of the triplet P-states can be reproduced
by L.S and tensor couplings, however, the long range potential has to be
a mixture between Lorentz vector and scalar, the latter being predominant.
In conclusion it might be said that it is quite astonishing how well the

simple potential models are able to explain or at least correlate the experi-
mental data. The predictions made and verified by experiments are particularly
impressive.
Of course, many detailed questions are still to be answered. It can be

hoped that more experimental and theoretical work will provide us with very
interesting information on the forces between two quarks. In particular more
knowledge on the long range binding potential will be valuable.
3. Hadronic and Radiative Decays of Charmonium
3.1 Possible decay modes of charmonium
The charm charge is expected to be conserved by the strong interaction

like isospin and strangeness. Consequently there are only 3 possibilities for
the charmonium system ce to decay.
a) The c and e quark separate and pick up another light quark pair to form
a pair of charm particles, e.g. ce + (cu) + (eu). The charm particle with
the lowest mass is the Do with m(D o) = 1864 MeV. The threshold for this decay
is therefore E = 3728 t~eV and the decays J/lJ; +00 and 1jJ' + DO are forbi dden
by energy conservation. The disintegration of higher states into charm
particles will be discussed in chap. 5.
b) Another possibility is that cc- annihilates \-lith the emission of gluons or
photons depending whether we are dealing with a strong or electromagnetic
decay. As in the case of positronium, a system with J = 0 can decay into
2 photons or gluons. The 2 photons can be real and such decays have been
observed e.g. X(2.83 GeV) +yy. The 2 gluons will be transformed into hadrons
and cannot be observed directly. However, for high mass states the 2 gluons are
expected to appear as 2 hadron jets.
A system with J = 1 couples to 1 virtual photon or 3 gluons (compare the
decay of ortho positronium into 3 real photons).
c) If the ce-system is not in its ground state it can cascade down by emitting
a photon or gluons with the ce-system staying together.
Diagrams for the various possibilities are shown in fig. 3.1a and b.
The ce states are bound by the exchange of many "soft" gluons. Since the energy
of the gluons is low their coupling to the quarks given by as(E) is large
(see 1.22).
If a virtual photon is emitted it can couple either to a lepton pair or a

j j
quark pa i r. These processes are proporti ona 1 to e 2_ as i nce the coupl i ng of
the photon is proportional to the charges of the particles attached to the
photon.
If gluons emerge from the cc annihilation they couple to light quarks. Each
of these gluons carries an appreciable fraction of the total energy ("hard
gluons") and hence the coupling as(E) is small. As a consequence these processes
are suppressed and the hinderance factor is determined by the number of exchanged
gluons (again in complete analogy to QED).
230 H. SCHOPPER
Hinderance
Example. factor
2
c
J~r } -e+ e~ 11+11-
:
c
:
~:: Ta
~
c
! : J/lil - ZIt,41t .... la
3
C
q~
lra
~ _
c _
c
c
0.-
J/IIIII--I-l1 c(O-+I+Y 3
a)
!
c
c
!
< l1clO+ -1- YY
Electromagnetic decays of bound charmonlum states

a
Example Hinderance
factor
~~\f
3
JlIII- 31t as
, '- S::.
c ,. __ ,.;/5
~ __ ' I
J/III-~ZIt a~
c ',~q
q
~' J/lil - Y+ 11
Y+ 11'
a a~
A'
C ;\'q
'0.
0 111'- J/lil +11
J/IIIZIt
a s3
b) HadronlC decays of bound charmonlum states
Fig. 3.1 a and b

3.2 Decays of J/~ and ~' (vector particles)
Vector particles with J PC = 1-- couple to the photon and hence can be
produced directly in e+e- annihilation (see Fig. 1.7b). As is well known the
IN, besides having been detected in p + Be .... e+e- + anything at BNL 25 )) has been
found at SPEAR 26 ) and soon after the ~,27). The decay of these particles was
investigated extensively at SPEAR 28 ) and at DORIS 29 ). Since detailed summaries 28 ,29)
have been published only the main points will be considered here.
~~~!_gg~gr~2~~~2Q~_Qf_~Q~~1_~~2_1gp~Q~2~_2g~~~_~22~b~
The width of the resonances turns out to be much narrower than the experimen-
tal resolution determined by the momentum smearing of the colliding e+e- beams
whi ch is about 1 to 2 MeV. But the true wi dths can be determi ned by a "tri ck" .
Assuming that the production cross section can be described by a Breit-Wigner
formula we have for the process e+e- .... J/~ .... final state
(3.1 )
where m is the J/~ mass, J its spin, ree' r f are partial and r the total
decay widths.
Integrating (3.1) over the energy E one obtains (with J 1)
(3.2)
If the three final states e+e-, ~+~- and hadrons are measured independently,
one can solve for the widths ree' r~~ and r = r h + ree + r~~. Since ree and
r~~ are found to be small compared to r h one has r h : r and from (3.2)
r (3.3)
ee
The integrals Lh and L (and L ) are found by integrating the experimental

ee w
cross section for the proper channel over the energy E. It is assumed that because
of the limited experimental resolution the resonance curve is widened but the area
is not changed.However, the integrated cross section has to be corrected for
232 H. SCHOPPER
radiative effects 29 ) which amounted to about 40 %. The results are shown in

table 3.1. Particularly striking are the narrow total widths, which indicate
the effectiveness of a new selection rule, i.e. charm charge conservation. As
explained above, the cc-system can only decay by higher order processes involving
hard gluons or photons and which therefore are hindered (see fig. 3.1).
Table 3.1 Resonance parameters of J/~ and ~'
lji/J lji'
SPEAR30 ) DORIS 31 ,32) ADONE SPEAR 27 ) DORIS 31 ,32)
m (MeV) 3095 4 *) 3096 2 3103 6 3684 5 3687 2
Lh (]Jb MeV) 10.4 1.5 9.7 1.2 9.6 1.7 3.7 0.6 3.06 0.34
L
ee (nb MeV) 790 965 141 790 200
L
]J]J (nb MeV) 870 100
r tot (keV) 69 15 87 20 67 25 228 56
ree (keV) 4.8 0.6 4.6 0.8 2.1 0.3 **)

r ]J]J (keV) 4.8 0.6 6.0 0.7 4.6 1.0
*) corrected value
'I'll') assuming ree = r]J]J
The most direct evidence for the photon-like quantum numbers of J/~ and ~'
is an o~servation of the interference between e+e- 7 y 7 ~ +~- and
e+e- 7 ~7 /~- . The cross section is given by
d 911 (1 2~ ree I 2
dGo -_ BtL + cos 28- ) I - j + u
=-m------,.-E-----,.ir~77'i"2 (3.4)
implying destructive interference below and constructive interference above the

resonance at m. The data are presented in fig. 3. 2 and clearly indicate the pre-
- E~pected Interference
o Interference
(0) o(3095)
b
.
'- 0.1
::I..
b::l..
0.01
3.000 3.090 3.095 3.100 3.105
0.15 (b) o(3684)
:: 0.10
b
'-
l
I::i 0.05
o L -__ ~ __ ~ ____L -_ _ ~ __ ~~
3.670 3.675 3.680 3.685 3.690 3.695

ENERGY Ec.m.(GeV)
Fig . 3 . 2
234 H. SCHOPPER
sence of the interference 28 ). Thus the assignment J PC = 1 for J/~ as well as

~' is proven. It is corroborated by measurements of the angular dependence of
e and ~ emission.
than 30 decay channels J/~ + hadrons have been observed. They will not
r~ore
be discussed here since an extensive review has been published previously28)

(Appendix 1).
The main conclusions drawn from the hadronic decays of the J/~ are the
following. The J/~ decays preferentially into final states with an odd number
of TI. The relation C = (-1) I . G together with G = C = -1 implies 28 ,29) iso-
spin IG = 0-, excluding I = 2 by the observed decay IN+pp. The fact that IN
also decays into an even number of pions violating isospin can be understood
quantitatively on the basis that the J/~ couples to the photon.
The isospin and G-parity of ~' can be inferred from the cascade decays
~' + IN + TITI and ~' + IN + n which account for 57 8 % of the ~' decays.
For example the experimental ratio (J/~oTIo) / (IN+TI+TI-) = 0.49 0.09 has to
be compared with the theoretical predictions 0.5, 0 and 2 for the isospin of the
TITI system I = 0,1 and 2, respectively. So clearly the pions have I = 0 and con-
sequently ~' and J/~ have the same isospin. The close similarity of the J/~ and
~' particles is obviously demonstrated by the fact that ~' + J/~ + TITI make up
about half of all the decays, whereas ~' + WTITI which has much more phase space is
about two orders of magnitude rarer. Also ~' + J/~ + TI o being I-forbidden is not
seen whereas ~'+ J/~ + n is allowed and was observed.
Finally it can be shown that J/~ and ~' behave as singlets with respect
to the approximate SU(3) symmetry of the 3 light quarks which is expected for
a charmonium state. The decays IN+ K+K- or KORo are forbidden for a SU(3)
singlet but allowed for an octet state. Indeed the experimentally observed
branching ratios for both the J/~ and ~' are very small (~ 10- 4 to 10- 3). The
SU(3) singlet nature can also be inferred from a comparison of J/W+ TIp and
KK*(892). In particular DASP results 34 ,29)indicate that the octet admixture
is very small.
Finally one might ask if we understand globally the decays of J/~ and ~'
or if major decay components are still unknown. By summing up all known decays
of J/~ and adding those channels which can be estimated by I-conservation one
arrives at about 70 % of the total decay width. It does not appear unreasonable
that the major part of the missing 30 % is due to decays involving n's i.e.
J/~ ~ n + anything about which very little experimental information is available.
Nevertheless a clarification of the situation would be welcome and in particular

some channels on the percent level can provide very interesting information
(see 3.24).
With respect to the ~' decays a little more than 50 % are associated with
cascade decays ~' ~ ~ + hadrons. The decays to intermediate P-states (see 3.3)
sum up to about 30 %. Adding the figure for electromagnetic decays (- 5 %) and
direct decays into hadrons (- 10 %) one arrives 28 ) at a total of about 95 12 %.
Not much room is left here for unknown channels, but again some rare undetected
decays might be of interest.
~;g~_Qi~~~~~iQ~_Qf_l~~!Q~i~_9~~~~~
A 1 -particle can convert to one virtual or 3 real photons. One real photon
is forbidden by momentum conservation and 2 photons by charge conjugation. Since
the transition via a virtual photon is of lower order it dominates.
The transition probability for (1 ) ~ y~ e+e- is for an s-state and a Cou-

lomb-like potential given by2,4)
(3.5)
where Q and m are the charge and mass of the quark involved and ~(O) is the
wave function at the origin (neglecting corrections as from annihilation channels).
Taking into account 35 ) "gluonic radiative corrections" equ. (3.5) has to be mul-
tiplied by {1 - 16 as / 31T}. For a general potential (see 1.23) I~(O) 12 has to be
replaced by m< dV/dr >.
In table 3.2 the experimental results of ree are shown together with ~(O)12
236 H. SCHOPPER
as derived 4 } according to equ. (3.5). For a "Coulomb" potential IljJ(O}12 is ex-

pected to risp - m3 , whereas for a linear potential - m. For the ground states
the rise is approximately -m 2, indicating that the effective potential is some-
where in between "Coulomb" and 1inear, in full agreement with the standard po-
tential (see 1.23). For a linear potential one expects that m <dV/dr> is the
same for the ground and the excited states. This is not true as a comparison for
+ -
Table 3.2: Vector meson decays into e e
Meson M (GeV) 02 r(keV} 11ji(0)12 (GeV) 3 x 10 2

p 0.77 1/2 6.5 0.8 0.29
w 0.78 1/18 0.76 0.17 0.31
fP 1.02 1/9 1.34 0.08 0.47
~/Iji 3.95 4/9 4.8 0.6 3.9
fi!' 3.68 4/9 2.1 0.3 2.4

~D1 3.77 4/9 0.37 0.09 0.44
fI!' , 4.15 4/9 1.8 to 3.3
tplll 4.41 4/9 0.44 0.14 0.72
ground states above the llne
ljJ, ljJ' and ljJ'" shows. A more detailed analysis for the excited states has been
carried out 6 ) for a harmonic potential and including Sand D-wave mixing.
The regularities found for the leptonic widths of the vector meson ground
states are useful in identifying new vector particles as the I (see chap. 6).
~~~~_Q~!:fQ~~igg~~_b~g~Q~if_9~f~~~
The Okubo-Zweig-lizuba 36 } rule claims that decays described by connected

quark lines are allowed whereas disconnected diagrams are suppressed.
u
allowed forbidden
Here the decay is shown as an example. The decay -+K+K- is allowed but
-+n+non- is associated to a disconnected diagram and hence suppressed. This
explains why the K+K- decay dominates in spite of its smaller phase space.
The functioning of the OZI-rule can be understood in terms of QCD. In

connected diagrams the interaction between quarks is provided by "soft" gluons
and hence is strong. In disconnected diagrams "hard" gluons have to be exchanged
withas(E) being small. If QCD perturbation theory is applicable, the rate of
a process involving n "hard" gluons, each carrying an energy E, -should be pro-
portional to [as(E~ n. We shall now see if this idea is compatible with ex-
periments.
~~~~!_~!~9!~_9!~~Q~~~~!~9_9!~9r~~~
A J = 1 state can decay to 3 real photons as in the case of ortho posi-
tronium. In analogy we expect that it can also decay into 3 gluons. (In the
electromagnetic case a transition to one virtual photon is also possible.
One virtual gluon is excluded, however, because of conservation of colour charge.)
Using the well-known expressions for positronium
electromagn. - 64.2 1
r( 1 -+ 3y) = 9 (,,-9) fii7 (3.6)
and replacing a by as' also applying a colour factor is (see chapt. 1.22) one
obtains 2)
strong (3.7)
This offers a possibility to determine as' Assuming that the conversion of the
3 gluons to ordinary hadrons goes practically with 100 % probability
f(l--+ 3g -+ hadrons) can be identified with the full hadronic width if connected
diagrams are forbidden by energy conservation. This is the case for J/~. In the
case of the only the decays into non-strange hadrons must be taken into account
and for the ~I the cascade and the radiative transitions have to be discarded
since they do not proceed via a 3 gluon intermediate state.
Instead of using the absolute values of f(l -+ 3g -+ hadron) it is expe-

dient to normalize them to the leptonic decay widths since then the unknown
wave function cancels. From (3.5) and (3.7) one obtains
238 H. SCHOPPER
r (1---+3~-+hadrons)_= 1410 (2/3)2 a 3 (3.8)

r (1-- -+e e-) ~ s
where Q is the quark charge.

If we now take the experimental data for ~, J/~ and ~', particles which are
almost pure cc states, one finds 4)
Table 3.3:
M (GeV) r h/ r ee as Remark
(jl 1.1 588 0.47 (jl -+- non-strange *J
hadrons
J/IjJ 3.1 14.4 0.22
1/J' 3.7 8 0.2 wi thout 1/J I -+- J/IjJ + ..

1/J'-+- Y + ...
The coupling constant a s indeed decreases with increasing energy as expected

in QeD (see 1.22) and is smaller than 1. Hence the application of first order
perturbation theory is justified at least qualitatively. Inserting as from
table 3.3 in (3.7) one derives r(J/~ -+ hadrons) = 72 KeV, which agrees quite
well with experiment.
In conclusion it can be stated that we understand in principle the narrow

widths of the bound charmonium states on the basis of charm charge conservation
and first order QeD perturbation theory. However, other models explaining the
OZI-rule could be as satisfactory. In the following and in particular in chapt.
3.3 we shall see that on the basis of QeD one can understand many
more experimental results than just the widths of J/~ and ~' .
Information on the inhibition by the exchange of hard gluons can also

be obtained by comparing inhibited processes to allowed ones, i.e. to decays
above threshold not forbidden by charm or strangeness conservation. Such
a comparison will be of particular interest for diagrams involving 2 and
3 gluons as for example~)
*) The reason for the small total width of ~ is that the OZI-allowed decay
~ -+ KK is suppressed by the small available phase space.
Table 3.4:
Forbidden decay Ratio 4) after number of exchanged expected inhibition

A11 owed decay phase space gluons factor
corr.
J/IjI ... hadrons - 0.002 3 a~ = 0.23 = 0.008
lji(4.4) ... hadrons
lji' ... J/IjI 1[1[ - 0.013 2 a~ = 0.22 = 0.04

p ... p 1[1[
<P ... p 1[ - 0.02 3 as3 = 0.47 3 ::: 0.1

<p ... KK
From this table one infers that indeed ~' ... J/~1[1[ involving only 2 gluons is less
suppressed than J/~ ... hadrons**). One cannot expect more than a very qualitative
agreement since the gluons in the two cases carry different energies and hence
as may differ. One further sees that <P ... P1[ is less suppressed thanJN ... hadrons
but the reason is quite different. Here it is the lower energy resulting in a
larger as that causes less inhibition.
*) It might be mentioned that ~' ... J/~ + 1[0 is forbidden by I-conservation

and has not been observed
**) The decay ~' ... IN +n has a large branching rati0 27 ) (4.3O.8)xlO- 2 in view
of the fact that it is p-wave and SU(3) forbidden, has very little phase space
(Q = 40_MeV) and is OZI suppressed. The large decay rate can only be explained
by a cc admixture to the n (see chap. 3.252) and hence no information on OZI
can be obtained.
240 H. SCHOPPER
~~~1~ __ g2~~l~_9!~~2~~~~!~9_9!~gr~~~
The decay of J/y permits even more detailed studies of the mechanism of the
OZI rule. Besides the singly disconnected diagrams discussed in the preceding
section there exist also doubly disconnected ones. In fig. 3.3 different diagrams
are shown which give rise to the decays J/y + w~~, wKK and J/y + ~~, KK.
The left and right columns contain the singly and doubly connected diagrams,
respectively. The two diagrams in the first line show the w~~ and ~~
decay without intermediate resonances, whereas the diagrams of the other two
lines involve the tensor (2++) particles f and f'. The f contains predominantly
u and d quarks and can therefore be connected to the w whereas the f' is made up
essentially of ss and hence prefers to decay to KK.
An interesting question is now whether the fourth gluon in the doubly con-
nected diagrams (not attached to J/ y) is also a "hard" one implying a suppression
a~ or whether it is comparatively soft yielding a hinderance - a s3 An analysis of
the experimental data with this aim is complicated, however, since the influence
of the intermediate resonances has to be determined. For this reason the overall
ratio 28 )
J/y + ~~ 0.21 0.09 ~ 1

J/y + wmi 1.0 0.3 "5 (3.9)
is not very conclusive. In order to reduce the influence of intermediate resonan-

ces, only events with M(~~) > 1 GeV were selected 38 ) and it seems that the suppres-
sion of the doubly disconnected ~~ decay is then stronger. Some interesting new
>
J/-==f:'~
'-' f_
r:;-K
K
Fig. 3.3: Singly and doubly disconnected quark diagrams.

information has recently become available from the PLUTO-collaboration 37 ).

They have studied the final state J/~ +rr+n+n-n-no and could observe the two
n -> ( wn +)n - -+- (n +n- n0 )n +nand
decays J1~-> B+- - J/1jJ-+-wf -+- (n +n- n0 ) (n - n+). The cor-
responding invariant or recoil mass peaks are shown in fig. 3.4. From these the
following numbers were deduced.
Final state Branching ratio r/eV

of IN decay %
n+n- n+n- n0 3.64 0.52 2548 360

wn+n- 0.78 0.16 546 110
wf 0.40 0.14 280 100
Bn 0.28 0.07 196 50
These results imply that most of the wn+n- decay width is due to resonance de-
cays. If these are subtracted the branching ratio for non-resonant wnn is of
the order of 0.1 % and hence comparable to nn. This is confirmed by selecting
only wnn events with M(nn) > 1.5 GeV, thus reducing resonance contributions, and
one finds B R(J/~nn) < 0.12 %. These data seem to indicate that the doubly
disconnected nn diagram is not much more suppressed than the simple disconnected
wnn.
This result is, however, in contradiction to measurements involving the f

and f' resonance. It was found 38 ) that the doubly disconnected decays J/~ -+- wf'
and f are about a factor of 10 rarer than the singly disconnected decays
J/~ -+- wf and f'.
More experimental work is needed to clarify the behaviour of doubly dis-

connected diagrams but certainly the decays of charmonium offer an excellent
tool to this end.
242 H.SCHOPPER
f PLUTO
~ recoil ing to '"

150
~1\
::!.
'"
'E
~10
50
o~~~~~~~~~
o 1.~ 1.0 H 30
Invariant MassM IGeVlcl) .
10
o0~5'-'-'-L.f1.;;'-"-'
0~15"""""""'1""'
.0 ...........U,H,........uJ
3.0
Invariant Mass M~eV/cl]
e+e -.. JI $ +8 +n- .. (Oln +),,- .. (n+n - no) ( n+n-)
.. ",f .. (,, +,, -,,0)( ,,+,, -)
Fig . 3.4
~~g?_~~9!~E!Y~_9~~~~~_Qf_~L~_~~9_~~_EQ_Qr2i~~r~_b~2rQ~~
In this section we shall discuss the radiative decays J/w ~ny, n'y, nOy
and fy. From these reactions interesting information on the admixture of cc-states
to the ordinary hadrons can be inferred. On the other hand radiative transitions
from J/~ and w' to other charmonium states (pseudoscalar and P-states) will be
dealt with in chap. 3.3 and 3.4. Here the interpretation will be quite different
since these transitions take place between practically pure cc states. In order
to be able to discuss the decays like ny, n'y a few ideas concerning the mixing
of quark states have to be recalled (3.252). The puzzle to be explained is the
experimental fact that the decays ny and n'y are more than an order of magnitude
more probable than JN~nOy.
~~g?!_~~E~r!~~~E~!_2~~~~_r~E~~_(~rl_~~r_E~~~!~2_
The decays J/w ~n~ n'y, nOy and fy are two-body decays and hence photons should
be monoenergetic. The search for narrow lines in the inclusive photon spectrum
at SPEAR was negative 39 ), but the first 3 decays could be detected at DORIS by
investigating the decays with 3 photons in the final state40 ,41), e.g.
J/w ~ ny ~ (yy)y but also in the decay42) J/w ~ yn' ~ y (ypo) ~ yyn+n-. The
corresponding invariant mass plots M(yy) show the nand n' peaks (fig. 3.5).
and M(n+n-y) sho\'1 (fig. 3.6) a clear n' peak provided f1(n+n-) is restricted
to the p region.
Very recently the decay IN ~ fy coul d be detected by PLUT0 43 ) by studyi ng

n+n-y final states, i.e. the decay IN ~ fy~ (n+n-)y. The invariant mass plot
M(n+n-) is shown in fig. 3.7. Besides the f peak also the p shows up. This peak
is attributed to the channel IN ~ pno~(n+n-) (yy) and a branching ratio of
(1.6O.4) x 10- 2 was inferred. Also DASp50) has observed the fy~n+n-y final states.
The results are summarized in the following table:
Table 3.5: Radiative decays of J/w

Final state ~ranCh~ng ratio r(eV) Experiment
x lO-l)
0.82 0.10 55 12 DASp40) 801

ny
1.30 0.4 87 27 DESY-Heidelberg 41 )
n'y 2.9 1.1 152 117 DASp 40 )
2.3 0.7 160 50 DESY-Heidelberg 41 ,42)
110y 0.073 0.047 5. 3.2 DASp40)
fOy 2.0 0.7 138 48 PLUT043 )
244 H. SCHOPPER
30 11
t
~20
::e:
-
c:.
'"on
C
.....~
10
11'
0.2 0.' 0.6 O.B 10 1.2

Myy(GeVlc/J
Lowest Photon Pair Mass
-- --OED 8r(l/lll -lIY):
_._.- QEO .Reflectlon from X (0 B2:0 1)'10"
- QEO.Retleclion fromX' lI Br(),\II - 1\'y),
(2.91!1l).10"
Fig . 3.5 DASP Preliminary Fig . 3.6 OESY -Heidelberg Preliminary

e+e- ~ yyy at J/ ~
J/ W(3 . 1) rr +n - yy
{ yy > 30
1241 events
J/~ ny
L.n
50.0
~o .o
~ 30.0
is
~ 20.0
...~ 10.0
0.00 0
O.S 1.0
M(n +n-) lC(vJ

Fig . 3.7 PLUTO Preliminary
IN fy n+n - y
prro n +n- yy
These figures are quite surprisi.ng at first sight. In terms of vector do-
minance one would expect that r(J/1jJ ->-YTIo) '" (a/y p 2 ) r(J/1jJ->- pOTI o )::: 1 eV with
r(J/w ->- pO,,0):::260 eV. This order of magnitude agrees with the experimental finding,
but why are r(J/w ->- n y) and r(J/w->-n'y) almost two orders of magnitude larger with
r(J/w->-n'y) ",r(J/w->- pOTI o)? And why is the ratio r(n'y) I r(ny) = 3.5 1.0 so
large?
A similar argument applies to the J/w ->- fOy decay which by vector dominance
is related to J/w->-fow with a measured 37 ,38) branching ratio of (0.40 0.14)x 10- 2
Hence the rate for the final state fOy should be comparable to TIOy but experi-
mentally it is much larger.
As will be discussed in the next section these anomalies can be understood in
terms of cc admixture to ordinary qq states.
Finally it should be mentioned that the corresponding decays w' ->- ny, etc.
have been looked for but not found 42 ,27).
~:~~~_I~~_~!~!~9_Qf_g9_~~~~~~
Thenyand y'n puzzle and some other experimental results (see chap. 3.3)
can only be explained in terms of mixing between cc states and qq states of
ordinary quarks. This mixing can be understood by extending the description of
the mixing of ordinary quarks.
It is well known that states of the ordinary vector and tensor mesons are
well segregated according to quark flavour. The neutral mass eigenstates exhi-
bit "ideal mixing" with (1/12~ ( uu ad) and 5S. As a consequence one expects
M(w) = t4(p) and M() = 2M(K*) - M(p). Both relations are fulfilled on the per-
cent 1evel .
The pseudoscalar mesons on the other hand show a strong mixing between
non-strange and strange quarks in spite of their different masses. As a conse-
quence the relations M(n) = M (lQ and ~1(n') = 2M(K) - M(TI) are strongly violated,
548 MeV against 138 MeV and 958 MeV against 854 MeV, respectively. The proper
masses are obtained with wave functions 44 )
::: -1 (uu - { + } .2
- + dd) ss
r,;!- (3.10)
/2'
246 H. SCHOPPER
which amusingly implies that the probability of finding a 5S or a uu/ad pair in the
nor n' meson is approximately equal.
The different mixing for the vector and pseudoscalar particles can be under-
stood 3 ,45,49) qualitatively in terms of gluon exchanges (see 1.22). To the mass
'matrix_correct~on termsAij have to be added Wh~Ch t~ke into account annihilation
terms qiqi ~ qjqj' If the 2 quarks are in a J = 1 state (vector particle) the
annihilation can only proceed via '3 gluons because of conservation laws. For a
JP = 0- state (pseudoscalar particle) annihilation via 2 gluons is possible. In
terms of QeD perturbation theory the first process is proportional to a~, the
latter to a~. For sufficiently large masses as < 1 and as a consequence the anni-
hilation corrections will be much more important for pseudoscalar particles than
for vector states, for which the Aij are negligible. Thus the different mixing
for the two kinds of particles can be explained in terms of QeD, but of course
one should not expect too precise results at low masses where as is still big.
If the charm quarks are included in the discussion the niass matrix for
pseudoscalar mesons can be written in lowest order perturbation theory:
M2 + AUU Aud Aus AUC

u
Aud Md+A dd Ads Adc
M- 2 (3.11)
qq AUS Ads ~12+A
s ss Asc
AUC Adc ASC t1 2+A

c cc
with M2u = Md = m2 'IT , ~12s = 2m K2_ m'IT 2' M2C = m2

nc
Isospin symmetry gives AUU = Add = Aud' Since Mc ~ 3 GeV one may assume
AI M2c 1 and in a first step to determine the A.. the annihilation q.q.+ cc
lJ 1 1+
may be neglected. With the experimental masses of 'IT, K,n and n' as input one can
solve equ. (3.11) and finds 46 )
AUU = 0.30, AUS = 0.21, ASS 0.12. (3.12)
SU(3) syr.lmetric annihilation \'Iould require AUU = >Us = \s and therefore it is

obvious that this symmetry is broken substantially. Also these findings can be
interpreted in terms of QeD. In lowest order perturbation theory one expects
(3.13)
The immediate consequence of equ. (1.1) should be "uu > "us > \s' as indeed is
found empirically. Furthermore, the elements "ij should satisfy the factorization
relation "iiO"jj = "ij 2. From (3.12) one finds ("uu "ss)1/2 = 0.19 in good agree-
ment with "us = 0.21.
As a next step one tries to predict the annihilation corrections for charm
quarks 46 ). From table 3.3 one takes a s (Ms2) / as (Mc 2) 2 which is also in
approximate agreement with equ. (1.1). Thus one obtains
(3.14)
Hence as expected the annihilation corrections for the charm quark are
smaller than for the ordinary ones and consequently M(n c ) ~ Mc to a very good
approximation.
The quark content of nc is given in lowest order by
(3.15)
nc = cc+~[us /" ss (uu + dd) + ssJ" sc / M2 (3.16)
and using (3.10)
n = cc + e: n + e: I n I (3.17)
c
with
We see that there is an admixture of ordinary quarks to the charm quark

state and again we notice SU(3) breaking since the admixture of strange and
non-strange quarks is different.
248 H. SCHOPPER
A similar approach for the vector mesons is possible in principle but

difficult in practice, since the relevant mass difference mw - mp is not
known precisely enough. If an equation analogous to (3.17) is written for
J/1jJ one has
(3.18)
Estimating the coefficients Ev and Ev ' from the decay of J/~ into ordi-
nary hadrons one finds E , E' ~ 10- 4 . As mentioned above such a small admixture
v v
for the vector charmonium states is exactly what one expects since in this case
the annihilation corrections are associated to 3 gluon exchange and hence
A~.
lJ
- ra
[s
(m.) . a (m .)] 3/2
ql s qJ (3.19)
implying that the A~j are about one order of magnitude smaller than the Aij for
the pseudoscalar states.
~~g~~_Ql~~~~~lQ~_Qf_r2~12~lY~_~~~2~~
The decay (1 -- ) ->- (0 -+ ) +y is associated with an electromagnetic Ml transition.
If the quark content of the initial and final state is the same the transition
probability is given by (see also fig. 3.8a):
r(1-- ->- 0-+) = j a (..9...)2 k3 n2

mq
(3.20)
where Q and mq are the quark charge and mass and k is the photon energy. The
overlap integral n is expected to be of order 1 if the transition takes place by
a spin flip without changing the other quantum numbers. Indeed for transitions
with ordinary quarks like w+rr0y, p+TIy, +ny, KO*->-KOy one finds values of
n between 0.6 and 0.9. Transitions with charm quarks in the initial and final
state like J/1J!->-nc + y will be discu~sed in chapter 3.3.
Let us now discuss the case with charm quarks in the initial and ordinary
quarks in the final states like J/~ ->- ny, n'y, nOy and fOy. Since the emission
of the photon changes only the spin of the system but not the quark content such
transitions cannot go via diagram a) of fig.3.8. A mixing of states has to
take place which can either happen after (Fig. 3.8b) or before (Fig. 3.8c)
photon emission. As a consequence the overlap integral n in (3.20) has to
be replaced by E n where E is one of the mixing parameters of equ. (3.17)
or (3.18).
a) q 1-
q
/Y O
- q
q
b) c
C iJj(n
4{~~
11 c (o-J 11 ,11 I (0")
Fig. 3.8
~~g~~!_I~~_~Yl_~~Y_~~Q_fY_fl~~!_~~~~~~
In the previous section arguments have been given that the admixture of
cp and w to J/1jJ is very small (, ' ::: 10- 4 ) whereas the admixture of n, n'
to nc is of the order of 10- 2. Hence diagram 46 ,49) c) in fig. 3.8 can be neg-
1ected and transitions I ike J/1jJ .... ny, n'y will go by diagram b). In this case
one expects from equ. (3.20):
(3.21)
with the , ' values given after (3.17). The experimental ratio (see Table 3.5)
is 2.9 0.8 in good agreement with (3.21). Thus the unexpected ratio between
these two decays becomes plausible.
With respect to the absolute rate equ. (3.20) yields
(3.22)
Simple calculation 46 ) based on harmonic oscillator wave functions \Ihose parameters

are adjusted to give the 1~(O)12 values as obtained from the leptonic widths
(see chap. 3.23) suggest n2 ~ 0.1. The overlap integral is relatively small be-
cause the transition energy is 1.5 GeV compared to decays of ordinary hadrons
like w .... ~oy, cp....ny where the transition energies are less than 400 MeV and
consequently n close to 1. As a result one expects according to (3.22) with
250 H. SCHOPPER
2 ~ 0.01 a width r(J/~ ~ ny) ~ 54 eV which agrees very well with the measured
width (see Table 3.5) of 55 12 eV. For JN~ n'y the corresponding values are
220 eV and 160 50 eV. These excellent agreements might be fort~itous in view
of the theoretical uncertainties but they indicate that lowest order QCD permits
one to understand the experimental results qualitatively.
The situation seems to be very similar for the JN~ fOy decay. Its bran-
ching ratio is of the same order as the JN~ fOw and hence vector dominance with
w ~ y as the source of the photons must be discarded, since it predicts a
fy/fw ratio of order ~y/y2w ~ 10 -3 . Again a diagram of the type b) in fig. 3.8,
where the photon is emitted by the cc must be dominant, this time leading to a
cc state with J PC = 2++ which then mixes with the f. A quantitative analysis 51 )
has recently been carried out. There it is shown th&t the angular distribution
of the photon contains interesting information.
~~g~~g __ I~~_9~~~_~L~:_~~y
For this decay the situation is reversed. Since nc - ~o mixing is forbidden
by isospin symmetry it is diagram c) of fig. 3.8 which is dominant and b) is
neglegible. Since the J/~ -w mixing is quite small, as explained in 3.2552, it
is plausible that the rate JN~~Oy is much smaller than JN~ny, n'y.
From eq. (3.20) one derives
2 __ r(JN~~O yI) (k2)2 nw~ (3.23)

- - . k.1 . n'J,~
~w r(w~~oY2) "'''
with nw~ ~ 1 and assuming n~~ = n~n = 0.3 (see 3.2531) one infers from the ex-
perimental values of the transition rates (Table 3.5) I~wl = 9 x 10- 4 . This is
indeed the order of magnitude expected from 3 gluon exchange (see 3.252). More
quantitatively one expects 46 )
= 41 (3.24)
The parameter ~ can be inferred from experiments in a similar way as ,J,'

",w 0 0 ",w
Analogous to (3.23) one has 24>W = r(4)~Yd f r(~ Y2) . (k2/kd 3 assuming
equal n. Inserting the experimental \'Iidths one obtains 1 I (5.4+0.9)' 10- 2.
This combined with the "experimental" I~) = 9x10- 4 yielt a ratio
4>w/~w ~ 61 which is again in reasonable agreement with the theoretical expec-
tation 41 of equ. (3.24).
In conclusion one might say that first order QeD calculations provide a
consistent picture for the radiative decays of charmonium into ordinary hadrons
explaining in particular the drastic difference between J/~ ~ ny, n'y and nOy.
A study of these decays gives simultanously a deeper insight into the mixing
of qq states.
Of course besides QeD there are other possibilities to understand the data.
Schemes involving SU(4) breaking have for example been developed 16 ,48) which do
not need an explicit potential. Many relations between various decay channels
could be deduced.
~~g~~~ __ Q~!_~~~~~1~12~~_~1~~2~~_~~_:_99_~~9_~~_~j_~_Q
If the transition rates are calculated49 ) taking into account only SU(3)
singlet and octet mixing for nand n' one obtains for all forbidden diagrams
with 3 gluon exchange (fig. 3.8c) for the final states the ratios
with the mixing angle 8 ~ 110. This is obviously in drastic disagreement with
experiment.
The 2 gluon diagram (fig. 3.8b) yields
Here the enhancement of ny and n'y with respect to nOy is properly reproduced
but the ratio r(n'y) / r(ny) cot 2 e ~ 30 is much too large compared to the
experimental value 3.1 1.
In conclusion it should be stated that ordinary all diagrams successful

in explaining the hadronic decays of J/~ and~' (see 3.24) have to be modified
by taking into account nc - n mixing in order to understand the radiative decays.
It seems that the decay ~' ~ J/~ + n is more analogous to the radiative
transitions than usual all forbidden decays. Its width 27) r (~' ~~n) = 9.3 1.6
keV is very large for its little phase space (Q = 40 MeV) and being p-wave, SU(3)
and all suppressed. Hence this decay seems to confirm the cc ++ qq mixing as
established in the radiative decays.
252 H. SCHOPPER
3.3 The pseudoscalar states
In terms of the charmonium picture one expects that the vector particles
J/~, ~' etc. with parallel quark spins (3 S1 states) are accompanied by pseudo-
scalar states with opposite quark spins (IS states) (see chap. 2.2). These
pc _+ 0
states have the quantum numbers J = 0 . Since they are analogous to the
n particle of the light quarks these states are usually denoted by nc (= 11So)
and nc ' (= 21 So ). Because of the hyperfine splitting one expects these states
to be somewhat below the J/~ and ~', respectively.
3.31 ~~E~~!~~~!~!_~~~~!!~_!Q~_!b~_01~~~~2_~~9_xl~~~~2_~!~!~~
Since the pseudoscalar states have even C they cannot be produced directly
in e+e- annihilation but can only be reached by decays from J/~ and ~'. Since
these decays are rather weak our experimental information on the nc and n~
states is very scanty compared to the rather complete knowledge of the vector
states J/~ and ~'. Indeed the existence of the nc and n~ has been in doubt for
quite some time. There is positive evidence now for nc which is identified with
the X(2.83) particle detected by DASP, whereas the existence of the x(3.45)
level still needs definite confirmation.
The possible transitions leading to nc and n~ are shown in fig. 3.9.
Fig. 3.9 Transitions involving nc and n~
A particle calledXwith a mass below IN has been detected by DASp 52 )

looking for 3 Yfinal states from the decay chain IN -+ yX-+y(yy). The invariant
mass plot for YY(high mass solution) is shown in fig. 3.10 and exhibits a clear
peak over the smoothly varying background 80 ) at a mass of (2.83 0.03) GeV/c 2
and a width of r = (29 14) MeV which is consistent with the experimental mass
resolution.
OASP
30
-VI
C
<1.1
>
W
I
'0
/ I
I
o 10 / I
z /
I
I
I
I
/
./
."
" ....
2.1 2.3 2.5 2.7 2.9 3.1

Myy{GeV fe 2 )
Highest Photon Pair Mass
QED
QED + Reflection from 1\ and 1\'
QED + Reflection from 1\ and 1\' + X ( 282)
Fig. 3.10
25.4 H. SCHOPPER
The branching ratio of the decay chain is given in Table 3.6. The DESY-Heidel-
berg 53 ) experiment sees also an excess of events in this mass region but be-
cause of their worse mass resolution they cannot give but an upper limit of the
branching ratio. The X particle has recently be seen in a pion-proton scattering
experiment at Serpukhov 56 ) and hence its existence seems to be well established.
Table 3.6 Branching ratios of electromagnetic transitions

involving pseudoscalar states nc and n~
Transition Product of br~nchi ng Experiment

J/1jJ ... n,. 1 ratios in 10-
IN ... y nc ... y(yy) (0.14 0.04) DASp52) 8o)
< 0.32 DESY-Heidelberg 53 )
IN -+ mc < 17 MPPSSSD 54 )
nc -+ yy > (8 3.5) calculated
J/~ -+ y nc -+ y(ypo)-+y(yrrrr) < 0.26 DESY-Heidelberg 53 )
J/~ -+ y nc ... y(pp) < 0.2 DASp 52 )
< 0.01 SLAC - L8L 55 )
~' -+ n" r DASp57) 80)
~' -+ y nc -+ y(yy) < 0.14
< 0.5 DESY-Heidelberg 53 )
~' -+ y nc < 11 SLAC - LBL 55 )
< 10 DASp57)
~' -+ yn
c
-+ y(ypo) < 0.18 DESY-Heidelberg 53 )
~, ... w nc -+ w(yy) < 0.23 -"-
~' -+ n~ I
~' -+ y n~ -+ y(yJ/~) 6 4 SLAC - LBL 55 ). PLUT0 55 )
~' ... y n~ ... y(yy) < 0.07 DASp57) 80)
~' -+ y n~ < 29 SLAC - LBL 55 )

n~ ... y J/lj! > 320 calculated
~' ... y n~ -+ y (ncrrrr) ... y(yyrrrr) < 0.06 DESY-Heidelberg 53 )
Because X + 2y the particle has even C and spin 1 is excluded. This is in

agreement with the expectation for nco
Themonoenergetic y-line associated to the allowed M1 transition

J/~ +nc has not been seen 54 ). Some indications were found 57 ) for
~' +nc + (yy)y but statistics are still too poor. The most puzzling fact is
the negative search for hadronic decays of the nco In particular upper limits
have been obtained for nc + pp and nc + ypo which are summarized in Table 3.6.
The transition probabilities will be discussed in chap. 3.33.
Evidence for a level at 3.45 GeV comes from an observation of the

cascade ~' + yx(3.45) and x(3.45) + yJ/~. This state has been observed with
the intermediate P-states (see chap. 3.4) but does not fit such an interpre-
tation. Hence the identification X(3.45)= n~ may be reasonable, although the
only information about its quantum numbers is even C, since it is reached by
an electromagnetic transition from ~'. Four cascade events have been found by
SLAC-LBL, one by DASP and three by PLUTO (see fig. 3.1~, some of which might
originate from background. Hence more experimental data are urgently needed
to establish this level.
The monoenergetic y-l i ne a ri sing from ~ I + n~ has not been seen (see
Table 3.6) as well as hadronic decays. The situation is similar to that for
the nco
~~~g __ Ib~Qr~~l~~!_~~~~~~~~lQ~_Qf_~r~~~l~lQ~_r~~~~
3.321 M1
- - transitions
- --- --
The decays J/~ncY and ~I+n~ are allowed magnetic dipole transitions.
The transition rate is given by
(3.27)
where Q and M are the c quark charge and mass, k is the photon energy and
n is the overlap integral of the initial and final wave function. Allowed M1
transitions are those between states which have essentially the same spatial
wave function and differ only in the spin state. Hence n~ 1 if spin-orbit
coupling and other spin-dependent effects are neglected.
With these assumptions one calculates 4 ,59) from (3.27) a decay width of
r(J/~+ync)~ 29 keV. From the branching ratio in table (3.6) the limit
256 H. SCHOPPER
r(J/~ -+ yn c } ~ 1.2 keY can be inferred, implying Q2 < 0.04.
Similarly one calculates 59 } the decay width r(~' -+ y n'} 17 keY whereas
c
the observed limit is r < 6 keY and hence Q2 < 0.3.
These comparatively small values of Q have caused some concern. Ml tran-

sitions between hyperfine partners should be rather insensitive to wave functions
and indeed this kind of calculation works to within factors of 2 or so for the
light mesons, where relativistic effects should make life much more difficult.
On the other hand, the large J/~ - nc and ~' - n~ splittings might indicate that
the Russel-Saunders approximation is poor. The discrepancy of a factor of 20
is certainly worrying.
The two "cross" transitions n~ -+y J/~ and ~' -+ y nc are forbidden Ml
since the main quantum number n changes from 2 to 1. Hence the wave functions
are orthogonal if spin-dependent forces are neglected. The width corresponding
to the second term in the power series of exp(ikr} is given by
r forb i dden (Ml) ::: 4~; -;.z- k7 Qforb (3.28)
where Qforb =< f I r21 i> and the meani ng of the other symbols is the same as in
(3.27) .
Using different wave func':ions r(~' -+y nc} between 1 keY and 10 keY was
found 4 ,5,14} which should be compared to the experimental partial width
r(~' -+y nc} ~ 2.5 keY. This agreement, however, has not much relevance. It
turns out that relativistic corrections to (3.28) are important 60 } and

factors of 10 are possible.
The situation for the decay n~ -+y IN is more compl icated. From the data
given in Table 3.6 one calculates a branching ratio B(n~ -+ y J/~} >(O.30.16)
which is very large. Since the total width of n~ is not known it is not possible
to give an experimental partial width corresponding to this branching ratio. On
the other hand the theoretical estimates deliver only r(n~ -+ yJ/~} and not the
branching ratio. Because of the unreliability of equ. (3.28) a better "theore-
tical" estimate can be obtained by assuming that the matrix elements of n~ -+y IN
and ~'+y nc are the same and correcting for phase space one obtains
r(n~ +y IN} :: (1/4}r W -+yn c ) 1/4) . 2,5 keY '" 0.6 keY.
In order to compare rand B it has been tried to estimate the hadronic

width r(n~ ~ hadrons) which should be practically equal to the total width.
As will be sho~m in 3.323 the hadronic width should be of the order of a fe~J
MeV. This yields a branching ratio B < 10- 3 in contrast to the experimental limit
B > 0.3.
Another way to compare experiment and theory is to consider the cascade

branching ratio B(lj!'~ n~Y~ J/1jJ yy) which experimentally is (6 4)x 10- 3 , whereas
theoretically one expects'" 3 x 10- 6 . This again is a large discrepancy; however,
the uncertainties for the calculated forbidden Ml transitions are very large.
3.322 nc ~ yy
- - - --
Transcribing the QED results for the singlet state of positronium one
finds for the decay rate 2 ,4)
(3.29)
Taking lj!(0) as determined for the J/lj! (chap. 3.23) one obtains r(n c ~ yy) '" 8 keV.
The experimental branching ratio B(nc~YY) > (8 3.5) x 10- 3 cannot be com-
pared to this expectation since the experimental total width of nc is not known.
Theoretical estimates will be discussed in the next section.
3.323 ~~9rQ~!_9~~~~_Qf_Dc_~~9_D~
Followin9 the analogy between QED and QeD the pseudoscalar nc and n~
can annihilate into 2 gluons besides into 2 photons. To derive the 2 gluon
annihilation rate from (3.29) one has to make the replacement 35 )
a 2 Q4 ~ 2a 2 s /9 and obtains
r(o- ~ hadrons) (3.30)
Inserting as = 0.2 and using lj!(0) as determined from the J/lj! one has
r(n c ~ hadrons) '" 6.4 MeV. This can be checked by an estimate which is independent
of the wave function and is obtained by taking the ratio of (3.30) and (3.7)
r(n c ~ g g) 27 7T
'" 100 (3.31)
r( J/1jJ ~ ggg )
258 H. SCHOPPER
Taking the measured width r(J/1jJ->- hadrons) = 69 keV one arrives at

r(Tl c ->- hadrons) '" 7 MeV in good agreement with the previous estimate. This total
width corresponds to a OZI hinderance factor of about ~ 50.
The total width of Tl~ can be estimated according to
r(Tl~ ->- hadrons) '" r(1jJ' ->- hadrons) ",/1jJ(0)

r(Tlc ->- hadrons) r(1jJ ->- hadrons) ~
/2
1jJ'
x I
~
M /2
,"
'I'
(3.32)
--.:.i
3.9 x
(l:l)
3.7 2 0 . 43
and hence r(Tl~ ->- hadrons) '" 3 MeV.
With this theoretical hadronic width one can now calculate the branching
ratios B(Tl c +yy) = r(Tlc+n) / r(Tl~ -> hadrons) '" 8 keV/7 MeV", 1.2 x 10 -3 . If
this is compared to the experimental 1imit B(Tl c -> yy) > (8 3.5) x 10- 3 one obtains
a factor> (7 3) of discrepancy. In view of the uncertainties this does not
seem too worrisome.
A more serious problem is the fact that so far no hadronic decays of either
the Tlc nor the Tl~ have been observed. One way out could be that these particles
decay to many different final states which are difficult to identify. With sta-
tistical model calculations some authors 61 ) have tried to estimate the proba-
bility of different final states. They find that indeed each channel contributes
only a few percent. e.g. Tlc ->- 2 TI + 2 TI - 2TI 0 7 to 15 %, Tlc ->- 2 TI + 2 TI - , 3TI+-
TI
2 to 5 % each, Tlc ->- pp + anything a few percent. The first channel which is do-
minant is hard to find. The limits given in table 3.6 may still be compatible
with these rates.
3.33 Q!~~~~~i9~_9f_Q~~~99~~~l~r_~!~!~~
The comparison between experimental results and theoretical expectations
concerning the Tlc = X(2.83) and Tl~ = x(3.45) particles can be summarized in the
following way:
a) Whereas the existence of the X(2.83) is well established, the x(3.45)
needs confirmation.
b) The large splitting between the 3S and IS states which originally has
caused much concern does not seem too much of a problem. It can be
understood quite coherently by attributing an anomalous gluon coupling
to the quarks (see chap. 2.2). Recently it has been shown 62 ) that the
existence of instantons may be responsible for the large splitting and
indeed it could be considered as direct evidence for the instantons.
c) Serious difficulties exist for the Ml transition rates. In particular
it seems difficult to explain the large discrepancies for the allowed
Ml transitions between HFS partners. The even larger discrepancies for
the forbi dden Ml "cross transitions" are perhaps 1ess worrying because
of the large theoretical uncertainties.
d) The real puzzle is the fact that no hadronic decays of nc and n~
have been seen, although partial width of several MeV are expected.
The hadronic transitions seem to be suppressed and one puzzling conse-
quence is the very large B(n~"'Y IN) > 32 %.
Because of these difficulties some authors have questioned whether the

X(2.83) and x(3.45) are really the IS states 87 ). Harari tried to identify the
x(3.45) with the 102 (J = 2-+) and Kr~semann and Krammer 63 ) interpreted this
state as a relativistic "time-like" P-state. Both proposals are not very
attractive since they do not solve the problems for the X particl.e and where
is then the n~ state?
Several authors l6 ,48,64) have shown that the transitions J/~ ... nc Y
and nc-+-rY can be hindered by SU(4) symmetry breaking. Indeed by a proper choice
of the parameters r(nc"'YY) < 3.7 keV can be obtained. Maybe relativistic
models are needed to interpret the decays involving nc and n~.
As a possibility to verify the nature of the X(2.83) and x(3.45) par-

ticles it has been suggested 65 ) to look for the decay ~' ... y X ... y(n+n-)X'" y(n+n-)(yy)
for which a branching ratio larger than 4 x 10- 5 has been estimated. Unfortu-
nately the upper limit given in table 3.6 has about this value and the hope that
this decay chain is strong does not seem to be realized.
In conclusion one has to state that more experimental information is

desperately required to solve the puzzle of the pseudoscalar charmonium states.
With all the success of the charm model this seems to be the only major trouble
that remains to be eliminated.
260 H. SCHOPPER
3.4 The intermediate P-states
According to the standard charmonium model one expects 3 triplet P-states

and one singlet P-state with masses between the J/~ and the ~'. As explained
in chap. 2 these P-states fall in between the two lowest 3 51 states because the
potential deviates from a pure Coulomb shape and hence the P-states are
lowered. Having C even the 3P-states cannot be produced directly in e+e- anni-
hilation but can be reached by an electromagnetic El transition from ~'. They
can decay either by a El transition to J/~ or into ordinary hadrons.
1 . Pc +-
The PI state wlth J = 1 cannot be reached from ~' by electromagnetic
transitions because of its negative charge conjugation. The decay
3P2 (2++) -+y IPI (1+-) is possible but phase space is probably very small. Hence
it is not surprising that the IP1 state has not been observed so far. The
following discussion has to be restricted therefore to the 3p states.
~~~!_~~e~~!~~~!~l_~~~~l!~_fQ~_~~:~!~!~~
The first evidence for an intermediate state called Pc was found by
DA5p ) in the cascade ~' -+ y Pc -+ y (y IN). Later this and cascades to other
66
intermediate states,which were given the generic name X} were seen at 5PEAR 54 ,55)
by DE5Y-Heidelberg 68 ) and PLUT0 69 ). The results are shown in fig. 3.11.
Clustering in the (IN y) invariant mass can be seen at 3.42,3.45,3.50 and 3.55
GeV. All these states have even C since they are reached by an electromagnetic
transition from a 1 state. The state at 3.45 GeV has a different character
than the others since its decay into hadrons has not been observed. As dis-
cussed in chap. 3.3 it is tempting to identify this state with the 2150 level
and since the relevant experimental data have been given in chap. 3.31 we
shall not discuss it here.
The mass distribution of the y J/~ system in the decay ~' -+ yyJ/~ as ob-
tained in recent measurements 68 )80) is shown in fig. 3.12 a and b. The peaks
at 3.51 and 3.55 GeV are clearly seen, there is a small indication at 3.41 GeV,
but no significant structure is observed at 3.45 GeV.
For the three P-states the monoenergetic lines corresponding to the transitions
~' -+y3 p have been observed (fig. 3.13) and the branching ratios were determined.
These branching ratios together with those for the yy cascade are collected in
table 3.7.
Ps IX -status
l.I6eV
lA t ~' -Xy
Ly~
" lIS6.v L II'II-
~.
I
U\ .8 llll6eV
}ll
9
4>< .I
8SP.. J!l.i itY
~
"0 .
",,~g I
lIS
PLUlij
~
g
0
,
0
x DASP
3.2 o SLiCILBL "
JI lIS 11 m
Mllyl (Gevl
Fig . 3. 11
DESY-Heidelberg
NoIS MeV
15 .0 r - - -- r - - - - r- - -r - - -- - ,
DASP ~' ~ yy J/ ~ Iji- yyJ/~
....
N
U
myy< .S GfN
:>-
OJ
708 ewnts
'"
~ 10.0
....
a
....'"c:
OJ
>
OJ
'::; S.O
o
z
~l (y IN)
Fig . 3. 12a Fig . 312b

262 H. SCHOPPER
300
200
100
)...
w
"):: 0
w
<I
0~
300
rf")
a:
w 200
a..
(f)
I-
z 100
:J
0
U
0
80 (e)
40
- 40
50 100 200 400 800 2000
Ey ( MeV)
Fig. 3.13
Single photon spectra from the decay of ~' as observed by MPPSSS054)
Table 3.7 ~,' -+ yP -+ yy IN cascades
State MeV+) B(ljI'-+y 3 p) B(ljI'+y 3 P+y(yJ/1jI)) Parameter++) a of Experiment

(%) (%) 1+ a cos 2 0
1 0.2 DASp 66 )SO)

3543 7 1.0 0.6 a = 0.3 0.4 SLAC-LBL 55 )
3561 7 7.0 2 2.2 1.0 MPPSSSD 54 )
2.3 0.6 DESY-Heidel. 6S )
1.1 0.3 DASp 66 )SO)

3504 5 2.4 O.S a = 0.1 0.4 SLAC-LBL
3511 7 7.1 1.9 5.0 1.5 MPPSSSD
3.3 O.S - 0.1 0.3 DESY-Heidel. 6S )
- 1.4 0.34 0.20 DASp SO )

3413 11 7.5 2.6 0.2 0.2 a=1.40.4 SLAC-LBL
3413 9 7.2 2.3 3.3 1.7 MPPSSSD
0.2 0.1 DESY-Heidel. 6S )
+) Mass from hadronic decays

++) For J = 0, a = I, for J = 1 and 2, a 1/3 and O.OS for pure dipole
transitions.
Information about the spin of a 3p state can be obtained from the angular
distribution of the first (low energy) y in the cascade with respect to the
beam axis. One expects a distribution proportional to 1 + a cos 2 e where e
is the angle between the first photon and the beam axis. With J = 0 for the
intermediate state one has a = 1. For J = 1 or 2 the prediction for a is not
unique since mixing of multi poles is possible. The experimental results for
a are also shown in Table 3.7.
No 3p -+ yy were observed 69 ).
264 H. SCHOPPER
The hadronic branching ratios of the

P-states have been determined by the SLAC-
LBL experiment. Invariant mass spectra for
different final states are shown in fig.
3.14 and the branching ratios have been
summarized by G. Goldhaber 7o ) (Table 3.8).
The most striking fact is that the 3.41
10
and 3.55 GeV states decay into n+n- and
10 K+K- whereas the 3.50 GeV level does not.
On the basis of the experimental ma-

terial presented in the previous section
one can try to determine the quantum num-
bers of the 3 levels with even C and see if
they are compatible with 3p states.
3.41 GeV: The observed decay into n+n- and

K+K- implies natural spin-parity
for this state, i.e. 0++, 2++ ...
30 3.1 ]I l6 H
Mass 16eV/c ' ) and it has to be an isoscalar. The
angular distribution is in agree-
Fig. 3.14 ment with J = 0 and hence one can
safely design this state to JPC=O++
3.51 GeV: The absence of n+n- and K+K- final
states indicates unnatural spin-
parity 0-, 1+, .... The two experi-
ments which measured the angular distribution disagree but both require
J F O. Hence the assignment J PC = 1++ seems plausible but needs confir-
mation.
3.55 GeV: The presence of n+n- or K+K- decays is again in favour of a 0++,
2++, ... state. The angular distribution indicates J F 0 and hence
the assignment J PC = 2++ seems plausible.
These assignments 71 ) are in perfect agreement with the expectation for 3p

levels. This ordering of the spins is further corroborated by the discussion of
the El transition probabilities (see next section).
The 3.45 GeV level which is the fourth intermediate state with even C
does not fit into the 3p level sequence and hence it is suggested to associate
Table 3.8 70)
x(3415) branching ratios
Decay Mode Events Efficiency B(l)!'-ryx) B(x->f) a)

B(X->f)
+ - (7 .52.1) x 10- 4 (1.00.3) x 10- 2
1T 1T 32 6 0.19
K+K- 27 5.5 0.16 (7 .82.3) x 10- 4 (1.00.3) x 10- 2
+ - + -
1T 1T 1T 1T 181 16 0.19 (3 .50. 7) x 10- 3 (4.60.9) x 10- 2
+ - + -
1T 1T KK 83 11 0.11 (2.80.7 x 10- 3 (3. 70.9) x 10- 2
+ - -
1T 1T pp (4.71.3) x 10- 4 (0.60.2) x 10- 2
23 6 0: 18
+ - + - + - (1.40.5) x 10- 3 -2
1T 1T 1T 1T 1T 1T 37 8 0.08 (1.90.7) x 10
1jJy 1 1 0.0011 (2 2 ) x 10- 3 (3 3 ) x 10- 2
a) We use B(1jJ'-ryX) = 0.075. The errors quoted for B(X->f) do not include the
overall scale uncertainty of 35 % due to the error ( 0.0026) in B(1jJ' -> YX).
X{3505) branching ratios

Decay Mode Events Efficiency B~ '->yx) B(X -> f) B(X -> f) a)
+ - + -
1T 1T 1T 1T 74 12 0.20 (1.40.4) x lO- J (2.00.6) x lO- z
+ - + - (0.80.3) x 10- 3 (1.10.4) x 10- 2
1T 1T K K 24 7 0.11
+ - (1.20.8) x 10- 4 (1.71.1) x 10- 3
1T 1T pp 6 4 0.19
+ - + - + -
1T 1T 1T 1T 1T 1T 48 15 0.08 ( 1. 90 . 7) x 10 - 3 (2.71.1) x 10- 2
1jJy 12 4 0.0011 (2.40.8) x 10- 2 (3411)xlO- 2
a) We 'use B(1jJ'-> YX) = 0.071. The errors quoted for B(X -> f) do not include the
overall scale uncertainty of 27 % due to the error ( 0.019) in B(l)!' ->YX).
X{3550) branching ratios
Decay Mode Events Efficiency B(l)!'->yx) B(X ->f) B(X -> f) a)

+ -
( 1.90.8) x 10- 4 (2.71.1) x 10- 3
1T 1T
o r 9 4 0.18
K+K-
+ - + -
1T 1T 1T 1T 89 12 0.20 (1. 70.4) x 10- 3 (2.40.6) x 10- 2
+ - + -
1T 1T K K 47 8 0.12 (1. 50.4) x 10- 3 (2.10.6) x 10- 2
+ - - (2.61.0) x 10- 4
1T 1T pp 13 5 0.19 (3.71.4) x 10- 3
+ - + - + -
1T 1T 1T 1T 1T 1T 23 15 0.08 (o.90.6) x 10- 3 (1. 30 .8) x 10- 2
1jJy 4 2 0.0009 (1. 00.6) x 10- 2 (14 8 ) x 10- 2
a) We use B(1jJ' ->YX) = 0.070. The errors quoted for B(X -> f) do not include the
overall scale uncertainty of 29 % due to the error ( 0.020) in B( l)!'-> yx).
266 H. SCHOPPER
it to the 21S level as discussed in chap. 3.3. However, since neither hadronic
decays nor angular distributions could be observed there is no direct experi-
mental information on the quantum numbers of this state except its even charge
conjugation.
~~~~_I~~Q~~~2~~!_~~e~~~~~2Q~_fQr_~!_~r~~~2~2Q~_r~~~~
The transitions from ~' to the 3p levels and from these to J/~ are electric
dipole transitions.
For ~'~ y 3pJ one finds 4 ) for the partial width
(3.33)
where Q is the quark charge, k the photon energy and J the spin of the final
state.
The experimental values for the branching ratio B (taken from table 3.7)
and the partial width r J (calculated with rtot(~I) = 228 keV) are shown
in table 3.9. The width r J for the 3 transitions are remarkably constant. This
is due to the circumstance that (2J + 1) k3 which according to (3.33) deter-
mines r J turns out to change only by a factor of 2 (see table 3.9) because
of the balancing between (2J + 1) and k3 The factor k3 alone changes by a
factor of 50. As a consequence the experimental constancy of r J supports very
strongly the assumed spin ordering since for different spins r J would change
by large factors. This statement is independent of detailed calculations of
the matrix element.
Original estimates 2 ,4) of the absolute decay rate were based on too
crude approximations and lack of precise knowledge of the transition energies.
The results of later calculations are given in table 3.9. They are in very
good agreement with the experimental r J . This confirms that the c-quark charge
is Q = 2/3 since for Q = -1/3 a discrepancy by a factor of 4 would emerge.
The transition probabilities for the decays 3PJ ~ yJ/~ are given by
(3.34)
Tab 1 e 3.9
3p
~ 7 Y J transitions
J k (MeV) (2J+1) k' B(I/I' +y3 pJ) % expo r J theoretical r J (keV)

(x 10 7 ) Jackson 4 ) Henriques 72 )
(keV)
2 123 0.93 7 2 16 5 19 14
1 173 1.55 7.11.9 16 4 27 21
0 271 2.00 7.52.6 17 6 31 22
Tab 1 e 3.10
J k (MeV) k'(x10 7) B(3 PJ+yJN) % theoretical r J (keV)

Jackson 4) 14
Eichten ) Henriques 72 ) Lane4)
2 466 10 14 8 300 400 367 320
1 416 7.2 34 11 230 300 258 200
0 318 3.2 3 3 100 140 120 90
268 H. SCHOPPER
Compared to (3.33) (2J + 1) has been replaced by 3 and hence no balancing of the
k3 dependence occurs. The experimental results for the branching ratios are given
in table 3.10. Since the total widths of the P-states are not known r J cannot be
determined. On the other hand, from (3.34) only r J can be calculated. As a conse-
quence a direct comparison between experimental and theoretical results is not
possible. Jackson?3) has tried to estimate the total widths from sum rules and
QeD but the results are not accurate enough to enable a detailed comparison.
Quantitatively one sees by inspecting table 3.10 that B(3 P1 .... YIN ) is lar-
ger than the other two branching ratios and also its absolute value of (34 11) %
is remarkably big. In contrast the theoretical value r J for this transition lies
between the other two. This indicates that the hadronic decays of 3P1 (3.51) are
suppressed compared to the other 23P-states. This can be understood in terms of
QCD as will be explained in the follwing section.
In terms of the simple charmonium picture the hadronic widths of the P-states
are given approximately by the annihilation of cc into two gluons 35 ) for J = 0,2
and into 3 gluons(or 1 gluon and a light qq pair?4))for J = 1.
(' ~ ~q
c c c
or
c c c
9
q
J =0,2 J =1
The 2 gluon annihilation can be calculated in a straightforward way using

the analogy to the 2 photon annihilation of positronium which gives:
(3.35)
where M is the c-quark mass, R' is the derivative of the radial part of the
wave function and the numerical constants are No = 256/3 and N2 = 1024/45.
The 2 gluon annihilation is obtained by the substitution a 2 Q4 .... a~(2/9) where
Q is the c-charge and one arrives at
1 R'(O) 12 (3.36 )
with N~ = 96 and N~ = 128/5.

The term IR'(O) 12 has to be calculated from an assumed potential and according
to whether a pure Coulomb, pure linear or a mixture of both is used, one
finds 4 ,35) IR'(O) 12 = 0.04 to 0.09 GeV 5 . From (3.36) one deduces then the
estimates
r(P 2(3.55j ~ hadron) 0.5 ~leV (3.37)

r(P o(3.41) ~ hadron) 2.0 MeV
Using the experimental PJ~yJN branching ratios one gets estimates for the
radiation widths of 70 and 60 keY, respectively. A comparison with the theore-
tical rJ(PJ~ y J/~) listed in table. 3.10 shows discrepancies of factors of 4
and 2. This can be considered as a reasonable although certainly not good
success of QCD.
The decay probabi 1iti es PJ ~ yy as cal cul ated from (3.35) are completely
negligible to the hadronic widths (3.37) and therefore also to the transitions
PJ ~ yJN . Indeed the two photOll decays have not been observed 57 ) with
B(~'~ PJy ~ yy y) : 4'10- 4 .
For the J = 1 P-states the QCD prediction is much less certain. It has
been argued 74 ) that the annihilation into g + qq dominates the 3P1 (J=I++)
decay whereas the ggg final state is most important for the IPI (J=I+-) state
which has not been observed so far. Both of these transition rates involve
logarithmic divergences at zero binding energy and
(:(3
r(P 1 ~ hadrons) = Nl ~ IR'(O) 12 ln (3.38)
with Nl = 128/3n for 1++ and 32019n for 1+-.

With the logarithm estimated as 2 In(llas) one finds for both J=1 states
r(P 1 ~ hadrons) = 0.15 MeV. The width is about 1/3 of r(P 2 ~ hadron) and the
reduction is mainly due to the extra power in as' This delivers also the
qualitative explanation for the large electromagnetic branching ratio of
P1(3.55) which arises from the suppressed 3 gluon annihilation of this P-state.
270 H. SCHOPPER
The general QeD formula (neglecting logarithmic divergencies) for the

annihilation of a state with orbital angular momentum into n gluons
may be quoted here 59 )
r '"
an
P (1T )2 ()
1R (0) 12 (3.39)
C C
where R() is the -th derivative of the wave function.
With respect to individual hadronic decay channels of the P-states very

little theoretical work has been done. The decay 3p ~ J/~ + rrrr is unimportant
and has not been observed in contrast to ~ ~ J/~ + rrrr which dominates. The
I
reason 5) is that the decays 3pJ ~ J/~ +rrrrhave little phase space and are hin-
dered by the centrifugal barrier.
3.5 Summary of charmonium spectroscopy
An enormous effort at SPEAR and DORIS has produced over the last few
years convincing evidence for the charmonium model and for QeD.
The level scheme of the cc system as deduced from experiments is shown

in fig. 3.15 including levels above the binding limit. All the various types
of levels have been found. The orbital momentum = 0 state with parallel spins
J = 1--, the triplet 3P-states and the lowest 3D-state. The - by now well
established - )(2.83) particle is very likely the lowest singlett IS-state
although its quantum numbers have not been determined experimentally. If the
state at 3.45 GeV exists and if it is the first excited IS-state remains to
be clarified. The singlet 1Pl-state is still missing but this is plausible
since it cannot be reached by radiative transitions from 1 states.
The spin-dependent splittings of the levels seem to pose no difficult-

ies in principle but rather yield valuable information on the spin-
dependent forces. The rather large splitting between the triplet and singlet
S-states may be associated either to anomalous gluon moments or the existence
of instantons. The splitting of the P-states indicates that the long range
confining potential contains Lorentz-scalar parts.
The hadronic and radiative transitions of the 1 states can be explained

qualitatively in the frame of QeD by the annihilation into hard gluons. In-
teresting information on the mechanism of OZI hinderance and the mixing of qq
states can be extracted.
The electromagnetic El transitions involving the 3p state are well under-

stood and contribute to the determination of their spins. The theoretical work
to explain the various hadronic decay channels is still rather scarce.
The Ml transitions involving the IS states present major difficulties.

The real puzzle seems to be the fact that no hadronic decays of the )(2.83)
and the X(3.45) have ever been seen.
Because of the exciting questions to be answered at higher energies the

clarification of many details of the charmonium system has been suspended . One
may hope that experimentalists and theorists will come back to these problems
since many interesting information on the interaction between 2 quarks might
be obtained.
IjI (4411
m:M$f~
FF*; OILOii*elc
t2
Mass.
GeV
18
3.6
34
12 21t or2K,4ltelc
3.0
2.8
t t
3P
t
30 1
1.2.3
Fig. 3.15
272 H. SCHOPPER
4. Total Cross Section and Inclusive Yields
A measurement of the total cross section e+e- + hadrons can give impor-
tant information on quarks and the strong interaction. As discussed in
chap. 1.13 this process can go via an intermediate resonance state with J = 1
provided the total e+e- energy E coincides with the rest mass of the reso-
nance.
4.1 Asymptotic limits of 0tot
Away from resonances the total hadronic cross section can be estimated
in the quark model on the basis of an asymptotically free theory with gluonic
corrections and one finds 76 )
(4.1)
where Q.1 are the quark charges, a s is the gluon-quark coupling constant whose
energy dependence is given by (1.1), and
o].l].l = (4n13) (a/E)2 = 86.8 nb I (E/GeV)2 (4.2)
is the e+e- + ].l+].l- total cross section. The sum has to be extended over those
quarks whose qq pairs can be produced at a given energy E. The factor 3 comes
from the 3 colour degrees of freedom (see chap. 1.21).
Since as(E) decreases with increasing energy E the ratio

+ -
R = o(e e + hadrons)/0].l].l tends to an asymptotic value from above. In the
energy region 3 to 8 GeV as is of the order 0.2 (see 3.21) and hence in (4.1)
the last term arising from gluonic corrections should be of the order of 10 %
or less. For the 3 light quarks the asymptotic value is Ru, d ,s = 3(2/9+419) = 2
and above the charm threshold one expects an increase by 6R c = 3(4/9) = 1.333
which would give R = 1013.
A heavy lepton with a sufficiently large mass can decay into hadrons and
hence can contribute to o(e+e- + hadrons). If the lepton is pointlike its
production cross section will be equal to 0].l].l (36 - 63 )/2 where the term mul-
tiplying 0].l].l gives the threshold behaviour for a spin 1/2 particle (6 = vic of
the 1epton). The contri but i on to R will be 6R l epton = B x (36 -6 3 ) I 2 where
B is the branching ratio of the lepton into hadrons. Since, however, in most
experiments measuring 0tot hadrons, electrons and muons are not identified decay
electrons and muons are taken for hadrons and consequently B = 1 and
6Rlepton ~ 1 well above threshold.
4.2 Experimental results for o(e +e- + hadrons)
The ratio R = (e+-

e + hadrons) / o(e +-e + ~ +-
~ ) in the range E = 2.4
to 7.7 GeV is shown in fig. 4.1, as measured at SPEAR 77 ). The data at lower
energies as obtained from Orsay, Frascati and Novosibirsk are still somewhat
scattered. In fig. 4.1 one notices a step around 4 GeV which is attributed
, ~~il l'I 'I '"I" \\,1

6
R
4
2
,11,/11"
0
0 2 4 ~ 6 7 8
Fig. 4.1 E c..... (G. V )
to the opening up of the charm threshold. The resonance peaks appearing around
4 GeV will be discussed In chap. 4.2.
In fig. 4.2 recent data from

DORIS obtained by the PLUT0 78 ,79)
6.0
and DASp 8o ) experiment are shown
R for the energy region 3.6 to 5.1 GeV.
The data of these two groups agree
1..0
very nicely and as far as the peaks
are concerned their position coin-
2.0 j.
cides with those of the SPEAR data.
However, as can be seen from
3.55 4.05 1..55 5.05
fig. 4.3, where the SLAC-LBL and
W (G eV j PLUTO data are compared, there is
Fig . 4.2
a discrepancy of 10 to 15 % at
energies below 4.5 GeV. This difference might be due to systematic effects,
274 H. SCHOPPER
e.g. acceptance corrections. These

should be smaller for PLUTO because 6
of its larger acceptance (86 %as com-

pared to 65 %for SLAC-LBL) and a more
general trigger. Recent data from
DELC0 86 ) seem to agree as far as the
absolute values are concerened with the 3 i
.J.
DORIS data but the systematic errors PLUTO
are still about 20 %. SLAC l BL
It seems important to clarify the

discrepancy between PLUTO and DASP
on one side and MARK 1 on the other,
since a 10 % difference corresponds Fig. 4.3
to half a unit of R which is quite rele-
vant for the interpretation.
The looser trigger enabled PLUTO to break down the total cross section
into charged mUltiplicities. as shown in fig. 4.4. It is remarkable that
10
2.0
1.0

h!!k/'t.J{T.IJ1r
I
0)0.2
a
3.5 1,,0 4,5 5.0
3,0
2,0 I
~Vt
iI\IlfII.1Ill I
1.0 ,/-
b) n.4
0
3.5 4,0 4,5 5.0
2D
1.5
..
....-,;~~r.I.
J __ _II
1,0
0.5 cJ n,,6
0
3,5 4,0 4,5 5,0
W [GeV!
Fi g. 4.4
these partial cross sections and in particular the 2-prong cross sections
show the same structures as 0tot' The dip at 4.3 GeV is deepest in 2-prongs,
and indeed at this energy R can be attributed almost solely to light quarks
and the heavy lepton.
In fig. 4.5 the various contributions to o(e+e- -+hadrons) are shown

schematically. For the 3 light quarks the asymptotic limit R = 2 is assumed.
It is remarkable that the new PLUTO results near E = 3.6 GeV are only about
10 % higher than this limit. The reason for this small difference can be due
to gluon corrections as discussed above or tails of lower resonances. The con-
tribution of the ,-lepton has been calculated for a mass 81 ) of 1.8 GeV. The
rise above the charm threshold can only be guessed but it was assumed that the
asymptotic limit of 10/3 has been reached around 5 GeV. The experimental va-
lue of R = 4.7 is higher than the expectation of about 4.2, which could be due
to tails of the resonances shown or because of the opening up of new thresh-
olds.
Fig. 4.5
In conclusion it can be stated that the data are in quite good agreement
with the 4-quark model and the existence of the heavy lepton and they confirm
expl icitly the colour factor 3 in equ. (4.1).
In passing it might be mentioned that the average number of charged

hadrons is given 29 ) by <nch> '" 2.1 0.7 ln (E/GeV) 2 with a value of about 4
in the region E = 3.6 to 5 GeV. It is remarkable that such a logarithmic rise
is found also in hadron-hadron collisions.
276 H. SCHOPPER
4.3 Unbound resonances
Above the threshold for two D-mesons Ethr = 2MD = 3.726 GeV charmonium
resonances can decay into DO. Above this threshold one expects therefore pro-
cesses of the type e+e- -+ ljJ* -
-+ DO according to the quark diagram
>------------~y-;--~--C~;=:=:: ~D
5
The total width of these unbound resonances should correspond to orcinary

hadronic widths in contrast to J/ljJ and ljJ' which are below the threshold and
hence can decay only by OZI hindered hard gluon intermediate states (see chap.
3.24). If the cc system has J PC = 1 these resonances should be observed
in o(e+e- -+ hadrons). However, a peak in this cross section does not necessarily
belong to a resonance. It could also be produced by the opening of a new threshold
and associated form factors which make the cross section drop fast at the high
energy side of the peak. Which case is realized has to be found out by detailed
studies. The situation of having discrete levels in the continuum is familiar
from nuclear and atomic physics and can occur if the potential is not Coulomb-
1 ike
4.31 I~~_=g{~~ZZl:~!~!~
A resonance very close to the charm threshold was found at (3.77 6) MeV
by the SLAC-LBL collaboration 84 ). Its widths was found to ber= (28 5) MeV.
This can be considered a beautiful confirmation of the arguments based on
charm and QCD since the ljJ',which lies only 88 MeV lower, has a much smaller width.
This demonstrates the difference between an allowed and a OZI hindered decay.
The mass, the total and leptonic widths have been predicted by the Cornell
theory group14,85) for the lowest 3D1-state with the help of the potential
model (see chap. 1.23) with astonishing accuracy. Therefore one might ask why
this resonance has not been detected before. The reason becomes obvious from
fig. 4.6a) which shows the uncorrected data. One notices that the resonance
falls on the large tail of the ljJ' and an appreciable amount of statistics had
to be accumulated.
In fig. 4.6b the results after applying

radiative corrections are shown. One
notices that the resonance shape is
10 asymmetric. If one fits the data by
t a)
a Breit-Wigner form one has to assume
that the width r(E) depends on the
as measured
energy. This in turn is expected be-
cause of the proximity of the DB thresh-
7
R old . With these assumptions good
fits to the data (see fig. 4.6b) can
t ,,+ ,ff+~+ be obtained. The resonance parameters

are given in table 4.1.
.++ +t+ + From the data to be discussed
rad corr b)
"*
later one calculates a DO threshold
at 3.87 GeV which is well above the
I~
resonance. As a consequence the 3.77
!! GeV particle should decay almost ex-
~, clusively into DB. The branching ratios
determined from the individual 0 decays
171 3 S~ lSS are listed in table 4.1. This property
makes the 3.77 resonance a clean source
Fi g. 4.6 for 0 particles whicn provides the
possibility for a detailed study of
these particles (see chap. 5.2).
With respect to the nature of the

3.77 level there seems to be no doubt that it should be identified with the
predicted 301 state. However, in a nonrelativistic treatment a O-state does
not couple to e+e-. It can attain a leptonic width by mixing with an 5-state
and since the ~' is so close one expects primarily a 1301-2 351 mixing. If one
writes 75 ) for the leptonic width
(4.3)
where a and b are mixing parameters and ree (3 01 ) is the leptonic width of
a pure 301 state. With relativistic corrections r ee (3 01 ) ~ 0.1 keV was found 75 )
278 H. SCHOPPER
and inserting the experimental ree for the 3.77 and 3.68 GeV states one
obtains a/{a+b) ~ 0.13. This implies an appreciable mixing. If the mixing is
expressed in terms of a mixing angle one finds 84 ) e = (23 3)0.
Table 4.1 Parameters for unbound resonances
State Mass r r
MeV/c 2 keY keY Branching ratios Experiment
30 3772 3 2B 5 0.37 0.09 0000 44 22 % SLAC-LBL82 )
1
0+0- 44 33 %
3770 6 24 5 0.18 0.06 OELC086 )
4035 2 55 5 0.7 0.1 PLUTO B3 )
4040 10 52 10 0.75 0.15 OASp 123 )
4146 4 47 11 0.4 0.1 PLUT083 )
4156 20 78 20 0.77 0.23 OASp 123 )
4414 5 33 10 0.44 0.14 SLAC-LBL 82 )
4400 3 33 9 0.3 0.1 PLUT083 )
4417 10 66 15 0.49 0.13 OASp 123 )
The total e+e- hadron cross section shows further peaks at 4.03, 4.15 and
4.40 GeV. If they are associated with resonances one can determine the parame-
ters given in table 4.1. These resonances could be 3S1 levels or 3D states with
some S-state mixing. Unfortunately there is no direct experimental evidence
to support such assignments. Also the predictions from the potential r,;odels
are not very reliable because of decay channels and other corrections (see
chap. 1.2). Structures can also be produced by opening thresholds, interfer-
ence effects between resonances and form factors in e+e- ~ DO, DO*, D* 0* etc.
Consequently the identification of structures becomes less reliable the higher
the energy.
The pronounced peak at 4.03 GeV is due to the conjunction of the 33S1 state
and the opening of the D*O* threshold at 4.012 GeV. That this peak is never-
theless a resonance can be seen in the Argand diagram for DO scattering I4 ,87).
The peak at 4.15 GeV could be attributed in the standard model to the 2301
state but it also lies just above the FF threshold at 4.06 GeV. To clarify the
character of this peak it would be interesti ng to look for F production which
is difficult, however. Nevertheless some evidence for F production has been
found 123 ) (see also 4.43). The Argand diagram supports the interpretation as
a resonance 14 ).
The resonance at 4.40 GeV finally could be associated 85 ) to the 43S1 state.
This is supported by the splitting 4.40 - 4.03 GeV ~ 400 MeV which agrees with
the expectation for 43S1 - 33S1 but does not fit so well with a O- state. A sim-
ilar argument applies to the leptonic width. The ratio ree (4.4) / ree (J/1jJ) ~ 0.1
seems to be compatible with an S-state but too large for a O-state (see chap.
3.21)85).
Some broad structure seems to become apparent around 3.95 GeV (see fig. 4.7).
This bump may have to do with the opening of the oD* threshold at 3.872 GeV and
with a zero in the 3S decay amplitudeS7 ). Similarly the broad structure around
4.35 GeV may be associated to the FF* threshold at 4.17 GeV. To check this it would
be very interesting to look for F and F* production in this energy region.
$L.AC-LBL SPI7
.0 DELCO Fig. 4.7

o DASP
~R in the region 3.9 to 4 GeV.
Because of the discrepancy in the
total height of R between SPEAR
and DORIS a "non-charm" contribution
.oR
of 2.5 units has been subtracted for
the SPEAR points and 1.5 units for
2
the OASP points.
O~L-------~~~9--------~.D~------~.J
WIGeV)
280 H. SCHOPPER
The situation in the 3.9 to 5 GeV energy region is by no means settled.

The threshold and the speculative assignments are shown in table 4.2. However,
the standard potential model (see chap. 2.1) may be wrong, in particular the
~-splitting (see chap. 6) seems to indicate that the confining potential is
not linear at higher energies. As a consequence the S-spacing could be much
smaller and all the structures above 4 GeV could be S-wave resonances. More
experimental and theoretical work is needed.
Table 4.2 Thresholds and speculative assignments to peaks

in the 3.8 to 5 GeV region
Threshold Pea k
Type E (GeV) E (GeV) Interpretation Observed

decays
DO 3.726
DO'" 3.872
- 3.95 threshold + form
factor
4.012
4.03 335
1
~O, DO"', o"ii*
+ threshold
FF 4.06
4.15 2301 Ft
or threshold
4.17
- 4.35 threshold?
4.28
4.40 Fr'"
The 4.03 GeV resonance is above the thresholds for DO,- DO-* and 0*'=0"* decay
but below the thresholds for decays into F-particles. At SPEAR a large sample
of data has been accumulated and the branching ratios for the 3 decay channels
(see chap. 5) have been measured 88 ,89). The results are shown in table 4.3. The
surprising result is that after correcting for the very different phase spaces
the 0* production is very much favoured. In a naive quark model the ratios of
the production probabilities should simply be given by statistical spin factors
which indeed favour the 0*5* channel but much less than experimentally found.
Table 4.3 Exclusive production of Do from decays of the

4.03 GeV resonance
DO from DO 50
D 5*0 +D*5 D* 5*0 Charged 0 Ref.
Fractions 0.05 0.03 0.38 0.08 0.40 0.10 0.16 0.11 88) 89)
experiment
Exp. after
phase space 0.2 0.1 4.0 0.8 128 40 70)
corrections
Statistical
spi n fac- 1 4 7 85) 23)
tors
Node in decay 0.04 4 30 85) 87)
amplitude
Oecoupling 0.9 4 175 90)
scheme
(normalized to second column)
The Cornell theory group has tried to explain the experimental ratios
in terms of a coupled channel model which produces a P-wave decay amplitude for
33S -+ (cu) + (cu) which is oscillating and has a node. As a consequence the cross
section a(e +e - -+ ~O)
-
has a zero near 4 GeV and the DO- production is strongly
suppressed with respect to O*D*.The ratios inferred from this model are
given (after phase space corrections) in the fourth line of table 4.3. The
282 H. SCHOPPER
tendency agrees with the experimental values, but there are still differences
of factors 4 to 5. While it seems possible to tune the model to get better
agreement on the ratios it seems difficult 88 ) to get reasonable ratios and at
the same time explain the maximum of R.
An alternative explanation 91 ) has been offered for the large D*O* pro-
duction by assuming that the 4.03 GeV particle is a D*5* 'molecule'. If this
were true, one expects by a rearranging of the 4 quarks the decay into J/~
which was estimated to occur on the 10 % level. PLUT0 92 ) has looked for the
inclusive production of J/~ in the energy region from 4.0 to 5.0 GeV. A
cross section of 31 21 pb has been found which corresponds to 0.13 % of
the total hadronic cross section. Consequently no enhancement due to 'mole-
cules' has been seen and the observed J/~ production can be understood 93 )
in terms of normal OZI forbidden transitions.
Still another mode1 90 ) to explain the enhanced D*6* production just

above threshold is based on a decoupling scheme. Its basic assumption which
can be defended on quite general grounds says that of a vector particle
decaying into 2 particles with spins J 1 and J 2 only the ground state and
the first J 1 + J 2 excitations (or daughters) can couple to these 2 particles.
Hence the ground state of a JP = 1- particle can decay into two J = 0 particles
but not the higher recurrences. This seems to hold for the p,wand Regge
trajectories separately (e.g. forbidding p'(1520) + 2n). If applied to char-
monium states, one expects the 13s (= J/~), 23S (= ~'), 33S (= 4.03 GeV)
states to couple to D*O* (each having spin 1 and hence J 1 + J 2 = 2) whereas
the 33S state should not couple to DO which explains the suppression. A de-
tailed analysis with form factors yields the ratios given in table 4.3 which
reproduce the tendency of the experimental values, but again some discre-
pancies remain.
This decoupling scheme could help to sort out the structures of the
total cross section. It predicts that the 43S state (perhaps at 4.40 GeV)
should not decay into D*5* pairs or any lower spin configuration like DO,
DD-* etc. The 23D and 33D should couple to D*-*
D whereas the 13D can couple
to DO (and higher spin configuration which are forbidden by energy conser-
vation, however). The 23D may also decay into D5* + OD* ~Ihich might help to
identify the 4.15 GeV peak.
4.4 Inclusive particle yields
Inclusive particle production can give interesting information in various

ways. Inclusive particle spectra are related to parton models and indeed the
momentum distribution of hadrons produced in e+e- annihilation 29 } found con-
siderable theoretical interest. However, it is beyond the scope of this review
to cover this subject. Here we are rather interested to learn something about
the production of new particles, which can produce steps in the total yield of
one kind of particle. As will be shown in chap. 5 the D mesons decay preferen-
tially to K mesons and also electrons. Therefore at the threshold for D pro-
duction one expects a sudden increase in the inclusive K and e production.
Simil arly the F mesons tend to decay to n and therefore a measurement of the
n yield seems interesting. Charmed baryons will lead to final states with
nucleon-antinucleon pairs, which have been looked for.
A step in K production at an energy ~Ihere the e+e- ... DD- channel opens
up provides very good evidence for the existence of charmed particles.
Strangely enough such an increase in the inclusive cross section could not
be seen for quite some time. Only at the beginning of 1977 was PLUT095} able
to see a step in the K~ cross section at an energy of 4 GeV. This was imme-
diately confirmed by DASp 96 } for charged K mesons (fig. 4.8). Assuming that
the increase found in the K cross
section has the same origin as the
step in 0tot one finds that 60 to 80 %
of the final states produced in the new
nb
.=ale+~ _K~ ...anythingl , PLUTO phenomenon contain kaons. The number
0: 1/2a(e"'e--K"! ...anything), DASP
of K~ per event is 0.39 0.06 and
since the number of all kaons is four
times larger one has 1.56 0.24 kaons
per event, which comes close to 2.
This figure is expected if pairs of
charm particles are produced and each
lO L5 5] decays into a kaon. Additional proof
Inclusive Cross Section of Kaons
that the kaons are associated to charm
Fig. 4.8 production comes from the K spectra.
284 H. SCHOPPER
In fig. 4.S two spectra for charged

kaons are shown as an example 96 ).
One spectrum was taken below, the
other just above the charm threshold.
The spectra agree above a K momentum
of about 1 GeV/c whereas at lower
momenta a clear difference can be
seen. This is exactly what one ex-
pects from D's decaying almost at
rest. Each D carries roughly the beam
energy and about half of thi s ener- 0.5 1.0 1.5
E IGeV)
gy or less is available for the kaon,
Fig. 4.9
i.e. EK <- Ebeam /2.
Kaon yields obtained 97 ,98) by ~1ARK 1 also show the pronounced rise at
4 GeV but the yields are larger than those found at DORIS (fig. 4.10). How-
ever, the MARK 1 data Ilere corrected by about 20 % for losses at low
K momenta and losses in the analysis. No such corrections have been applied
to the PLUTO data. The remaining discrepancy has about the same size as the
difference of the total cross sections, except at an energy of 4.415 GeV.
Here the MARK 1 point lies significantly higher than the PLUTO result. This
could be due to the fact that the MARK 1 data were taken exactly at the peak
of the resonance whereas the PLUTO data are averaged over 15 MeV.
PlUTO Kt
OSUC-l8l ..;
K'
Eo. [GoY]
Fig. 4.10
Charmed particles can decay by weak interaction to final states con-

taining only hadrons or also leptons (semileptonic decays) (see chap. 5).
As a consequence one expects a step in the cross section for electron
production. Indeed since ordinary hadrons have very small branching ratios
for the decay into electrons, the cross section for electron production
should jump from practically zero to a large value at charm threshold. The
electrons originating from charm decays are accompanied by several hadrons
since e.g. e+e- -+ DD- -+ (D-+ K. e \I)+(D-+
-
hadron). The decay of a heavy lepton
on the other hand leads preferentially to 2 prong final states, e.g.
e+e- -+ TT-+ (T-+e \!\!)+(T-+j.J\!\!). Therefore the electron yield is measured
for events containing 2 or more additional prongs besides the identified
electron.
Three groups have measured inclusive electron spectra. DASp Bo )126) and
B6
DELCO ) use Cerenkov counters for electron identification, whereas I1ARK 1 98)
has recently be complemented by a lead glass wall. The results of the three
groups are presented in fig. 4.11. Indeed, one notices the rise of
Re = a(e +e- -+ e + anything) / a(e +e- -+j.J +-
j.J )
DELCO
R from zero below charm threshold to values
.~
of the order of 0.2 above 4.0 GeV. A peak

can also be seen at the 3.772 GeV resonance.
There is a marked difference between the
DASP and DELCO data around 4.2 GeV. The
0..1 dip seen by DELCO at 4.25 GeV is not pre-
sent in the DASP data but this may be due
to the averaging of the DASP data over
(),J
300 MeV. The dip may even be consistent
with no charm production since the sepa-
1.1 II ~tj
ration of the lepton decays was not per-
fect and hence the value of Re at the dip
could be entirely associated to T decays.
# ----- ---------
_ .,i....
..
06''- .... ".
. It should be noted that also the total

cross section has a dip at that energy
(see 4.2).
Fig. 4.11
286 H. SCHOPPER
One can now attempt to calculate the branching ratio Be of the charmed
mesons into electrons:
r R'
e (4.4)
r 2 Rcharm
R~ is derived from Re as shown in fig. 4.11 by correcting for losses of electrons

at low momenta and of events with 2 charged tracks and by subtracting the con-
tributions from T decay (10 to 20 %). Furthermore Rcharm = R - RT - Rold ' where
ROld is the R value below charm threshold.
The results are shown in fig. 4.12. The SPEAR data tend to be lower than
the DASP results which is mainly due to the difference in R used to calcu-
1ate Be'
Fig. 4.12 demonstrates that Be does not change much with energy. This is re-
markable since at higher energies F mesons and charmed baryons can contribute.
The constancy of Be can either mean that F and baryon production is negligible
or that the branching ratios of these particles are similar to those of the
D meson.
~ 0.30,..,---,--,--.---.,---.-.--.--.-01""
-E
b 0.25
u
~ 0.20
l-
n.
~g 0.15
-oe{
~~
f/)(!) 0.10
w~
(!):I:
~!:!! 0.05
Woe{
~!f5 0.0'--'----.J_----1_--L._~_...i..._...L.._..L-_-41-..IJ
3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 7.0
Ec.m.(GeV)
Fig. 4.12
Kaons in the final state indicate the decays of D mesons, the appearance
of n is a sign of F decays (see chap. 5). The inclusive yield of n offers
therefore an interesting possibility to find energies ~Ihere F production is
appreciable.
Unfortunately the signature identifying an n decaying into 2y is not very

strong and therefore it seemed for sometime very unlikely to find them at all.
A solution to this problem was found by the DASp8o)99)123) collaboration.
Ins tead of looking for the reaction e+e- -> FF- they tri ed to observe e+e- -> FF-* .
The F* decays F* -> yF and with an estimated photon energy of about 100 ~'eV
(see chap. 2.22) this low energy photon yields an additional signature. With
F -> nTI -> yyTI one arrives at final states containing 3 photons, 2 originating
from the n and one having low energy. The invariant mass distributions for
two photons are shown l23 ) for different energies in fig. 4.13a and b. For the
energy interval 4.36 to 4.48 GeV a clear n signal can be seen besides a TI o
peak. For the other energy intervals nonsignal appears. If the back-
ground is subtracted one obtains the inclusive n yield as a function of the
e+e- energy, as presented in fig. 4.14.-The cross section is consistent with
zero except at 4.4 GeV, where almost all of the n mesons are accompanied by
a low-energy photon. A rough estimate gives a value 0 <n > - few nb, where
n n
o is the production cross section and <n > the average multiplicity. These
n n
data indicate that n production is strong at the 4.4 GeV resonance but small
everywhere else. This might indicate a similar behaviour as for D production
where DO* production is dominant somewhat above threshold (see chap. 4.33).
Exactly because of this analogy the F search was started around Ecm::: 4.4 GeV.
Recently the analysis of events containing n could be improved 123 ) and

as a result the requirement of having a low-energy photon in the final state
could be dropped. Hence a search for n originating from e+e- ->FF became possible.
A clear n signal was found in the energy interval 4.10 to 4.22 GeV (fig. 4.15)
which is just above the threshold at 4.06 GeV. No n's were observed in the
energy interval 4.00 to 4.06 GeV as is expected. This F production just above
threshold can be compared to D production at 3.77 GeV and should be an ideal
source to study F decays. Indeed electrons indicating semileptonic decays have
been observed (fig. 4.15).
288 H. SCHOPPER
Events with low E. Y
4.36 s W s 4.48 GeV

40 e+e- - VYYlow
lt + (i!: 2 charged) + X
30
~
-
~
~
Kl
c
c
w 20
10
o 0.5 to
Myy in GeV
Fig. 4.13a
Events with low E.y
1,0
1,.00 O!E W O!E 4.10 GeV
>QJ
~
Lf)
10
.......
....~
c
w
10
0.50
Fig. 4.13b
290 H. SCHOPPER
0.4 ,---,--,----,--,----,---,-----.
-g 03
1
;>-0.2
F
f
~0.1
-----l------r-----------+
I:>
4.0 4.2 4.4 4.6 4.8 5.0 5.2
Fig. 4.14
4.10 :5 W :5 4.22 GeV

1.0
30
>QJ
:L
oLf)
JJ
.~ 20
C
UJ
10
Myy in GeV
Fig. 4.15
4.44 Inclusive Po' antinucleon and strange particle production

--------------------------------------------------------------
Inclusive no production has been reported by PLUT0 79 ). The Po spectrum
follows the same exponential shape as found for charged pions but the absolute
cross section is about a factor of 2 higher. The yield as a function of the e+e-
energy is shown in fig. 4.16. The large error bars indicate systematic
PLUTO
RI,I 01..- --so.anll
a,.,. prelinVnary
t t tt
l
w
3.S 4.0 4.S 50 GoV
4.16
errors which are independent of energy. The data, therefore, exhibit a step
just below 4 GeV, the threshold of charm, but also of heavy lepton production.
The average value of R(po) = 1.3 for W> 4 GeV may be used to estimate which
fraction of charged pions comes from vector mesons. The result is that more
than 50 % of all pions originate from vector mesons. Predominance of vector
mesons over pseudoscalars is expected from the quark model simply on the basis
of statistical spin factors.
The observation of antinucleons and strange baryons is very interesting

since it can indicate the production of charmed baryons. Some early results
for p production have been obtained by DASp 1oo ) (fig. 4.17). Very nice results
from the MARK 1 detector have b~en published recently101) (fig. 4.18) for p
and A + Aproduction. The antiprotons were identified by TOF measurements.
The difference of R(p + p) = 2R(p) between the DASP and MARK 1 data is probably
due to different cut-offs at low p momenta. The data indicate a rise by a fac-
tor of about 2 in the region between 4 and 5 GeV.
292 H. SCHOPPER
Rp-=~
""----.-----.-----r-----.
0.20
0.15
0.10
t
005 f /\cAc
OL-__~____~II _____~I~
3.5 4 4.5 5
Vs(GeV)
Fig. 4.17
0.6 r- a) + +++++
+
IQ. 0.4
....,.. +
~0.2
a
0.10 b) t+
~+tit t
1<
+
<
a:: 0.05
a 1
3 4 5 6 7 8
Ec.m(GeV)
Fig. 4.18
The increase of the cross sections coincides with the expected thresholds
for singly charmed baryons (strangeness 0 or 1) around 4.4 GeV. If the in-
crease is indeed due to charmed baryon production, the ratio of charmed bar-
yons to uncharmed baryon production is about the same as the corresponding
ratios for mesons.
R{A+A) is about 10 to 15 % of R{p + p) 2R{p). The smallness of this

figure indicates that the weak decays of charmed baryons prefer modes with
nucleon plus strange meson and pions in the final state instead of A's.
The production of antisigmas has been measured by the UCLA-SLAC colla-

boration 89 ). The ~ were identified by combining the ii ~lith a n in a mass
plot. The ii were detected through their annihilation properties, and their
momentum was measured by TOF. At an e+e- energy of 7 GeV 27 events were ob-
served, but practically none at 4 GeV. This gives an increase of
ClR{l: ) = 0.11 0.05.
+ -
The most direct evidence for the production of charmed baryons in e e
annihilation would be the observation of a peak in the mass distribution of
the expected decay particles. This has not been achieved so far and a compa-
rison with the two charmed baryon candidates found in photo 102 ) and neu-
trino 103 ) production is not yet possible. Here is a rich yet uncovered field
for e+e- experimentation, although it seems that baryon production cross
sections are in general relatively small.
294 H. SCHOPPER
5. Charmed Mesons and their Weak Decays
As we saw in preceding chapters, the charm hypothesis 106 ,107) can explain
most of the experimental findings connected with the cc-system. The final proof
for charm comes from the detection of mesons consisting of a charm quark and a
light antiquark and their antiparticles. Of course, one expects that according
to the two relative spin orientations each cq state occurs as para (J = 0) and
ortho (5 = 1) particle~ The following nomenclature is generally accepted:
cs-
C +1 : D+ cd, DO -
cu, F+ pseudoscalars
C -1: D cd, DO -cu, F -cs J = 0
DH cd, D*o cu, FH cs vectors

etc. J =1
The D+ and DO form an isodoublet, whereas F+ is an isosinglet (see chap. l.21) .
The masses of these particles were predicted 3 ,23,34) with astonishing

accuracy on the basis of the naive model (chap. 2). The experimental masses
are collected in Table lb. The great triumph of the charm model is the existence
of all the cq states that had been predicted, at least for the states where
c and q have orbital angular momentum zero. P-states for the cu and cd system
have been predicted with the following masses 23 ,24):
D(11 P1 ) 2.5 GeV

D(1 3Po ) ~ 2.4 GeV
D(1 3P1) ~ 2.6 GeV
D(1 3P2 ) 2.6 GeV
None os these states has been identified so far.
The excited states can decay to the ground states by hadronic or electro-
magnetic interaction, e.g. D* -+ D + 1T, D* -+ D +yor F* -+ F + y(see chap. 5.4).
The higher states are therefore expected to have rather large widths. The
ground states (predicted to be the pseudoscalar states), however, can decay
only by weak interaction and hence should be quite narrow. This is because the
strong interaction conserves charm charge and therefore a cq system is stable,
whereas the weak interaction can convert a c quark to light quark.
In order to discuss the weak decays of the D and F mesons we shall very
briefly recall the structure of weak interactions in its simplest form.
5.1 Minimal theory of weak interactions
The minimal formalism 109 ,110,111) of weak interaction is based on SU(2)w

doublets. Assuming that there are 4 leptons and 4 quarks one has
Q Leptons Quarks Q
o 2/3
- 1 -1/3
All these doublets contain left-handed particles. The right-handed quarks and
leptons are singlets under SU(2)w. (For larger groups see for example 108 ) and
chap. 6.)
Each doublet gives rise to a current q1 y~ 1+l

5 q2=(q1q2)
- or -~lY~1+l5 ~2=(~1~2)
-
where q and ~ stand for a quark and lepton, respectively and 1 and 2
belong to the same doublet. In the following we shall use the abbreviation
(q1q2) and (~1~2). A V-A coupling is assumed.
The d and s quarks are eigenstates of the strong interaction. This need
not be true for the weak interaction. Hence the most general possibility is
the Cabibbo structure, which allows mixing between isodoublets. As can be shown
the most general case for 4 quarks is covered by the rotation 108 )
d ' d cos8 + s . s i n8
(5.1 )
SI -d s i n8 + s cos8
where 8 is the Cabibbo angle.

A mixing for the leptons has no physical meaning if the neutrino masses are
zero, since in this case the rotated as well as the unrotated states are
eigenvalues of the mass matrix. The mixing between quarks ensures that there
is only one conserved baryon number whereas the non-mixing of neutrinos results
in separately conserved electron and muon lepton charges.
296 H. SCHOPPER
The effective weak Hamiltonian in lowest order is given by

4G
Hwea k -72
_ F ( J ~ + J h ) ( J ~+h
J) t (5.2)
The lepton current J~ and the hadron current J h can both be split in a charged
current JC and a neutral current IN with
(5.3)
J~ (Lid') + (cs') (5.4)

cosG(ud + cs) + sinG (us - cd)
JNh uu + cc - d'd' - s's - sin 2 GWJ e~

(5.5)
uu + cc - dd - ss - sin 2 G W Je~
In (5.5) the electromagnetic current Je~ has been added and the Weinberg angle 111 )
Gw is the SU(2)W x U( 1) mixing angle.We shall not be concerned here with the uni-
fication of electromagnetic and weak interactions but for completeness the
following relations may be quoted:
4G F 2
-~ M - 37.3 GeV
72 - 2M 2 'W- sinGw '
W (5.6)
e/9 = sin GW' MZ = MW/cos GW
where g is the SU(2) coupling constant, and MW' MZ are the masses of the
charged and neutral intermediate boson.
More important for the following discussion is the GIM mechanism 107 ).
The existence of the charm quark had been postulated in order to explain the
absence of strangeness changing neutral currents as in the decays KO ~ ~+~
or K ~ n v v. The hadronic neutral current (5.5) is invariant under rotations
in the Cabibbo angle, as can be seen explicitly. J~ does not contain G and
in particular the terms sinG. cosG (as + sd) are cancelled. Hence transitions
between the d and s quarks are forbidden and the missing of the 16s1 = 1
K decays is thus explained. However, the GIM mechanism automatically forbids
a1so the neutral transiti ons between u and c quarks. The experimental evi-
dence for this expectation will be discussed later.
The charged hadronic current (5.4) contains two parts. One is proportional
to cosG involving the transitions u++d and c++s. The other contains the factor
sinG associated to the transitions u++ sand c++:I. From the hinderance of
strange particle decays relative to non-strange decays one finds sinG ~ 0.2.
Hence.one calls the decays proportional to cosG Cabibbo allowed and those in-
volving sinG Cabibbo forbidden.
The behaviour of the hadronic currents is nicely summarized in fig. 5.1.

Each corner of the square represents a quark whose charge is indicated in brackets.
Q.
U(IIJ) _s~~L"'.QL s(-h)
cosec"O.97
=lllQI=l
--lllQI=O
d (-I/J) --:Sin9c--- e(IIJ)
Fig. 5.1
The full arrows show the Cabibbo allowed, the broken arrows the Cabibbo for-
bidden decays. Both involve a change of the electric charge by one unit.
Neutral current transitions would have to be represented by diagonal

arrows. The transition d++s has [lIS[ = 1 and the transition u ++c has [lIC[ = l.
These strangness or charm changing neutral currents are cancelled by the GIM
mechanism since the amplitude going from s to d via u is proportional to sinG'
cosG whereas via c a factor -sinG. cosG is picked up and these two contributions
cancel each other. Similarly the transition u++c via sand d is cancelled. The
only neutral currents which exist connect each corner of the square to itself,
e.g. u++u.
It might be mentioned that the c quark was not only invented for the
GIM mechanism but also establishes symmetry bet~leen quarks and leptons. One
consequence is that the sum of all fermion charges is zero (counting the quark
charged three times because of colour) ensuring the absence of anomalies as
required in renorma1izab1e theories.
298 H. SCHOPPER
The most important message we get from this structure of weak inter-
actions for the detection of charmed particles is the strong coupling between
c and s, resulting in a predominance of strange particles in D and F decays.
More specifically one has the diagrams
K- 1];T]', KK Tt
u 5 5 5 U d
---w --w ___ w
u c 5 c U c
D F+ D
Cabibbo allowed (- cose) forbidden (-sine)
where the W boson can couple on the ri ght side to 1eptons (semil epton i c
decays) or quarks (hadronic decays). In both cases DO decays preferentially
into K and F into n. n'. Therefore these strange particles in the final state
are important signatures to observe the D and F.
From (5.4) it can be seen that the Cabibbo allowed decay transforms
c++s whereas c++d is suppressed by tge in the amplitude. Hence one expects
the following decays with leptons only in the final state
The decay F+~ ~+v is analogous to Hence one

~ 112)
expects for its decay rate
(5.7)
Assuming that the decay constants fF and fK are equal, one finds with mF = 2.03 GeV
r(F-+jJv):< 4.2 x 109 s -1 . (5.8)
The actual value could be larger by a factor of - 2 since one expects

> >
fF - fK - fn and indeed fK :< 1.28 f n
As will be shown later the decay width into hadrons is of the order
rtot(F):< 10 13 s -1 and hence the branching ratio 8(F-+jJv) ~ 10 -3 to 10 -4 which
is very small. However, the jJ spectrum is characteristic for a two-body decay
and might be observable.
The decay F+ -+ e+ ve is suppressed relative to (5.7) by a factor (m/mjJ) 2

and is negligible.
The particularly interesting decay F+ -+"[ v "[ is unfortunately very difficult

to observe.
The leptonic decays of 0 particles are Cabibbo forbidden and hence are re-
tarded by a factor tg 2 e ~ 0.05 with respect to allowed decays. As a conse-
quence their branching ratios are very small.
In summary it can be stated that leptonic decays of 0 and F are very weak
and indeed have not been observed so far.
Such decays are of particular interest since they offer the cleanest
way to study the charm current. This is because they originate from the product
J; J h (see 5.2) which implies that the hadronic current appears only once and
is multiplied by the well-known leptonic current. The two possible decay modes
are:
decay amplitude LIS III

c -+ s R,+v cose 1 0 lIQ = tiC = LIS
+e
c -+ d R, ve -sine 0 1/2 lIQ = lIC = 111
300 H. SCHOPPER
If the c quark is combined either with u or 5 one has the following

decays:
K- 1I III I n-
U 5 5 5 U d
---(' 1+
--<~ 1+
--(' 1+
U C S c IT c
D F+ D
allowed forbidden
The decay rates can be estimated using symmetry arguments and making assumptions
about the form factors l12 ,l14). One finds for example l12 ,l14)
-
r(D +Kev)
+ ~
( mD
5
_+
mK ) cot 2 e . r(K +n e v) (5.9)
+ -0 + 11 -1
'" r( D +K R- v) '" 1,4 x 10 s (5.10)
and with
r(Do+ K*-R-+v) / r(Do+K-R-+v) '" 0.54 (5.11)
one obtains
(5.12)
For the decays of the F meson one can derive similar expressions but they
are less reliable because of nand n' mixing. Estimates are l12 ,l14)
0.9 x 1011 5- 1
'" 1.1 x 1011 s-l (5.13)
0.3 x 1011 s-l
implying
r(F + n'R- ve ) / r(F + n R-v) = 0.29 (5.14)
The decay F + w R- v is OZI suppressed since the w has very little strange
quark content.
For the Cabibbo forbidden decay one finds the estimate l13 )
r(D o -+7fQ,v)/r(D
- 0 -
-+KQ,v)::: 2tg 2 8:::0.1 (5.15)
and similar suppression factors for D+ -+7foQ,+v and F+ -+ KQ,+v. The decay
F+ -+ 7foQ,+v is Cabibbo and OZI suppressed and hence negl igible.
Besides the transition rates one can calculate the momentum spectra of
the leptons. These will be discussed below together with the experimental
results.
Measurements of the ratios given above present an important test of the

structure of the hadronic current. Significant violations could indicate new
pieces in the hadronic currents.
The understanding of nonleptonic decays is complicated by the influence

of the strong interaction on the weak current product J~ J~t. Charmed and strange
particle decays are affected differentlyl12):
1) asymptotic freedom suggests a reduced influence on charmed particle

decay amplitudes
2) the quark field of strange particles contains the combination cc - uu
which couples to a gluon. In the Cabibbo-allowed decays of charmed par-
ticles, on the other hand, a corresponding coupling does not exist. For
Cabibbo suppressed decays the combination ss - dd has a small effective
coupling to gluons.
3) two-body decays of charmed particles are more energetic and therefore
less influenced by final state interactions.
5.131 g~~~~_9~~2~~_fQ~_!~Q:~Q9~_~Q~!~Q!Q~1~_9~~~~~
Because of the arguments given above two-body decays of charmed particle
should be particularly suited to be described in terms of simple quark dia-
grams l12 ,114). Amplitudes are usually calculated using colour factors and
exact SU(3) and ignoring gluons. This approach should allow to test the
general structure of weak interactions involving c quarks. Corrections should
be small compared to ratios of decay rates.
302 H. SCHOPPER
5.1311. ~a~i~b2 ~llo~e9 ~o~l~lo~i~ ie~ays

The graphs are shown in fig. 5.2. Examples are the decays
0-+ + 0 + + + .
o ~ K ~ , D ~ K ~ and F ~ n~ . The dlagrams on the right of fig. 5.2
are suppressed by a colour factor 1/9 compared to those on the left. This
is because of the sum over the three colours in the outgoing meson for the
diagrams on the left, whereas the colours must match in the diagrams on the
right side and hence the colour enhancement is lacking. The decays of 0+ are
K-,K*; ...
5
+ p+
/~"'"
_-L..-/d
u c
K, KilO
d 5
y:,P ...
+ +
--.L.-/a
4>,lJ, ...
S 5
5 C
Fig. 5.2: Cabibbo allowed two-body decays of charmed particles
special since the two diagrams lead to the same final states and therefore
add coherently.
One immediately finds many relations for the Cabibbo allowed decays:
2r (F+~ n~+) = (3/4)2 r(o+ ~ KO~+) =
(5.16)
18r (Do ~ KO ~o) = 9r (F+ ~ K+ KO )
One interesting consequence is the suppression of F+ + K+K o which makes this

decay unfavourable to look for F+.
Decay rates for Cabibbo allowed decays have been calculated taking into
account short distance gluon corrections, SU(3) breaking and form factors l12 ).
The results are given in table 5.1.
Table 5.1: Theoretical nonleptonic 2-body rates l12 )

of Cabibbo allowed decays of D and F
Decay mode Decay wi dth Decay mode Decay width

1011 s -1 1011 s-l
colour enhanced
DO + K- p+ 2.3 DO ->- j(*o p+ 3.4
K*- p+ 1.7 KO p+ 3.7
K- 'IT + 1.7 -*0
K 'IT + 2.7
K*- 'IT + 0.96 KO 'IT + 3.7
colour suppressed colour enhanced

DO ->- po -*0K 0.14 F+ ->- q,p+ 2.2
wo -*0K 0.14 np + 1.6
'IT o -*0
K 0.22 n'p+ 0.5
n K*o 0.06 q, 'IT + 1.1
n' R*o 0.0005 n 'IT + 1.1
pO ~ 0.09 n' 'IT + 0.6
w j(D 0.08
'ITo KO 0.2 colour suppressed
n RO 0.08 F+ ->- KH K*o 0.4
n' j> 0.03 K+ K*o 0.5
K** KO 0.2
K+ KO 0.4
Besides the graphs shown in fig. 5.2 other graphs can contribute to
two-body decays which are of two types
304 H. SCHOPPER
>--~:
C
F-poTt+
d
Fig. 5.3
The amplitudes for these graphs are multiplied by the factors mn2 / (m
and mK / (mo - mK), respectively,and hence they can be neglected relative
2
m~) r-
to the graphs of fig. 5.2, The expectation
(5.17)
can be considered a significant test of the ideas underlying the quark graphs.
If an additional qq pair is added to the final state of the left graph one
obtains a prediction for the 3-body decay F+ ~ n+n+n-
(5.18)
which provides a similar test as (5.17).
5.1312 ~aeiebQ forbidge~ ~o~l~!o~i~e~a~s
In nonleptonic decays the Cabibbo suppression proportional to sine

can occur on either or both vertices. This is sho\</n in fig. 5.4.
~>---< :>---( ~>---{ j--<~

6. S 1 0 0 -1
/::, I 1 ill. 1/2, 3/2 0,1
Amplitude - cos 2 0 cos0. sin0 -sin0' cos0 -sin 2 e
Example D .. K-n+ D .. nn' D .. TT+TT- D0 .. KOno
Fig. 5.4: Cabibbo forbidden nonleptonic decays

The corresponding single or double suppression factors are tg 2 e ~ 0.05 and

tg~e , respecticely. From the graphs the following relations can be decuded l15 ):
r(O o -+ 1T+1T- ) 2 r(o+ -+ 1T+1TO) = 1. r(F+ -+ K0 1T+) = 2 r(F+ -+ K+1TO) (5.19)

r(Oo -+ K-1T+) r(o+ -+ K0 1T+) 2 r(F+ -+ n 1T+) r(F+ -+ KOK+)
All these ratios are of order tg 2 e.

For the singly suppressed decay F+'-+ nK+ the decay rate is estimated l12 ) as
r(F + -+n K+ ) = 0.05 x 10 11 s -1 .
The part of the interaction that gives rise to nonleptonic decays is

given by J~ J~t of (5.4). This part of the Hamiltonian transforms as a sum
of the symmetric ~O and ~ representations of SU(4). Broken down into SU(3)
representations one has
SU(4) SU(3
L'lC = 0 IL'lcl 1
20
~
84
., (L'l1
8,27
1/2)
(L'l1 = 3/2)
~,
15,
-
15
6
--- ,.....~ ,...." "...J
The decay F+ -+ 1T0 ~ +V is Cabibbo and OZI suppressed.
According to the well known L'l1 = 1/2 rule which was found experimentally,
the nonleptonic decays of strange particles (L'lC = 0) are enhanced. From the
above table it is seen that this octet enhancement can be interpreted as 2o-plet
enhancement in an exact SU(4). If this were true the octet enhancement should be
associated to a sextet enhancement of the decays of charmed particles. Whether
this concept is true is one of the interesting questions that can be answered
from a study of nonleptonic decays of charm particles.
Of course, the predominance of the 2o-plet with respect to the 84-plet

can be due either to an enhancement of the 2o-plet or a suppression of the
84-plet. An enhancement of the 2o-plet and consequently of the sextet would
lead to a large nonleptonic decay rate and a small semileptonic branching ratio
306 H. SCHOPPER
for the D decays. The suppression of 84 and hence of 15 forbids the decay
D+ ~ KO~+. As will be discussed below the semileptonic branching ratio for D
decay is 11 3 %,which is not small ,and also the D+'" KO~+ decays have been
observed. In conclusion it seems that there is no sextet predominance and a
different explanation for the octet enhancement has to be found.
Decays of D and F mesons in many hadrons which are Cabibbo allowed are
proportional to cos 4 e and have ~C = ~S = 1. They can be written symbolically
DO ... (K + n 'IT)0
D+ ~ (K + n ~)+ (5.20)
F+ .... (11 + n ~) +, (11' + n ~ )+. (K K + n ~ )+
An example is the decay

D+ .... K- ~
+ ~
+
quarks: cd su, dU, du
with C = I, S =0 C = 0, S -1, Q = +1
Such a final state having Q = +1, S = -1 is called "exotic" because it cannot be

formed from an s quark and one u or d in the framework of SU(3) but rather needs
3 quarks. These exotic states with a strange particle are a signature of charmed
particle decays. The decay to a nonexotic state (total quark content su) like
.... +
~ ~
su, du, ua
is unfavoured for charm particles, but allowed for normal particles like
K*+ ~ K+~+~-.
The inclusive leptonic yield and the total hadronic width can be estimated I14 ,116)
by comparing the graphs
=(:: r
From the first two graphs one concludes
f(charm -+Ji.v + hadrons) xcos 2 0Xf(1J -+ ew) 0.7xlQ12 s-l

(5.21)
where m and m are the c quark and IJ mass, respectively and

c IJ 1
f (IJ -+ ew) = 0.45 X10 6 s - .
Assuming that the final state interaction of the quarks indicated by the
bubble in the diagram on the right goes with unit probability one finds
f(charm -+ hadrons) = 3 (:c) 5x cos 2 0 x f(IJ-+ ew) = 2.1 x 10 12 s-l (5.22)

IJ
where the factor 3 comes from the fact that ua occurs in three colours.
If, however, a sextet enhancement as discussed in 5.132 exists f(charm -+ hadrons)
would increase by a factor of 20. If (5.21) and (5.22) are applied to D decay
one expects a leptonic branching ratio
~ ~ 30 % without sextet enhancement (5.23)

~1.5 % with sextet enhancement
This provides an excellent possibility to test experimentally the existence

of such an enhancement.
If the lifetime of a DO is sufficiently long it could mix with Do by

first-order jt-C! = 2 neutral currents in a similar way as KO and 'K 0 mix by
!t-S! = 2 currents 117 ). If this ~Jere true nearly half of the decays of DO would
308 H. SCHOPPER
decay as if it were a Do, e.g. to K+n- instead of K-n+. Another way to look for
DO - DO mixing is in associated production of DO, DDN etc., where the kaon
originating from the two charmed particles have opposite charge for no mixing
(S conservation) and equal charge for mixing.
Mixing must compete with Cabibbo forbidden decays with ~C = -~S at a

fractional level of order sin'8 - 10- 3 (see chap. 5.1312). A way to discrimi-
nate l15 ) between the two possibilities is through production and decay of
lji( 3722)
(5.24)
The final state with two equally charged K's arises through both mechanisms,
whereas ee solely through mixing provided the semileptonic ~Q = ~C rule
is valid.
Since DO's decay promptly in-l0 -13 s there is little time for mixing and
hence the theoretical predictions for mixing are very small (much less than
a percent; for literature see l13 ).
~~~ __ ~~E~rj~~~!~1_r~~~1!~_2~_g_9~~~~~
The decisive confirmation for the charm idea introduced to explain the
J/ljiparticle and the other charmonium states is the existence of the charmed
mesons DO, D+ and F+ (see 1.21). Hence it is not surprising that a big
effort was started very soon at SPE.L\R and at DORIS to find these particles.
Initially thi~ search was unsuccessful l18 ) and also the expected rise in the
inclusive K yield at charm threshold was not seen (see 4.41). The experimental
work at SPEAR and DORIS was complementary since MARK 1 concentrated on the
investigation of hadronic decays, whereas PLUTO and DASP were looking for
semileptonic decays. Both kinds of searches became successful in spring 1976
after K identification became possible at MP.RK 1 by time-of-fl ight measure-
ments and electron identification was introduced at DASP by Cerenkov coun-
ters. PLUTO could establish for the first time e-K o correlations.
The observation of hadronic final states has the advantage that in-
variant and recoil masses can be measured. Hence beautiful and very precise
mass values could be published by the SPEAR groups. This is not possible
for the semileptonic decays since the neutrino remains undetected. However,
interesting information on the weak interaction could be extracted from the

observation of these decays by the DORIS groups.
The experimental results will not be described in their historical order

but rather in a more systematic manner. We shall start with D meson production
at the 3772 MeV resonance. This energy is below the threshold for D+ production
and hence the decay of the 3772 MeV resonance provides a pure sample of D mesons.
The decay 1/1(3.77 GeV) + DO is very close to threshold which allo~ls a very
precise determination of the D masses. The mass is calculated from m = (E2_p2)1/2
where the energy E of the D must equal Eb, the energy of the incident beams.
Eb is known very preci sely and has a spread of only about 1 MeV. Being close to
threshold,p2 ~ 0.08 (GeV/c)2 is small, and thus the effect of any error in p on the
mass is minimized. As a result a mass resolution of about 3 MeV/c 2 can
be obtained, which is better than normal mass resolutions by a factor of
5 to 10 (see chap. 1.13).
The data were collected 104 ) with MARK 1. Charged kaons are identified
by time-of-flight measurements and neutral kaons by measuring the dipion mass
and checking the consistency of the vertex position with the kaon decay time.
The invariant mass spectra for

different K~ combinations are shown
D' Di
6lJ ~(Ol (dl in fig. 5.5. Three clear mass peaks
('It' Kslt'
10 10 are seen in neutral final states and
20 two decay modes for charged final
states. With an additional lead glass
:::!.
Kt It, It counter also the decay into K+- ~-+ ~ 0
10
with direct observation of the ~o
)0
could be seen l19 ).
20
The results on the masses are
~1(Do) = 1863.3 0.9 MeV/c 2
0
til; til 1.85 19O
0
til; til 1.15 t90 M(D+) = 1868.3 0.9 MeV/c 2 and
MASS IGeVlc 21 M(D+) = M(D o) = 5.0 0.8 t~eV/c2
Fig . 5.5
310 H. SCHOPPER
The mass difference is known more pre-

cisely since several systematic errors
cancel.
The angular distribution of D's

relative to the incident beams must 30 (a)
t
be of the form P(S) - (1+ a cos 2 8),
lal ~ 1 for any 0 spin and a = -1
for spin O. Fig. 5.6 shows the angular 20
distribution for 0+ and Do decays.
The values of a are a = -1.04 0.10
and -1.00 0.09, respectively, con- 10
sistent with the spin 0 assignment
for the 0 mesons.
o ~~------~------~
In order to verify that a Do 20 ot ... K+ lit lit (b)
is produced together with a Do the
+ -
recoil mass against the K- rr+ system
restricted to the Do mass has been
determined. The result is shown in 10
fig. 5.6~. A clear peak at the mass
of the Do is seen. Thus associated
production is verified which proves O ~~----~ ______~J
the conservation of the new quantum
-1.0 o 1.0
number charm. cos 0
Fi g. 5.6
An important point is that for
o mesons originating from the 3.77 GeV
branching ratios B can be deter-
mined, whereas at higher energies
only a . B can be given.
This requires, however, two assumptions: the ~(3.77) must have a definite
isospin (0 or 1) and its only sUbstantial decay mode is ~O. The results ob-
tained under these assumptions are shown in table 5.2.
Table 5.2 : D branching fractions
Mod e Branching fraction Reference

in %
DO - +
-> K 7f 2.2 0.6 104
-0 + -
K 7f 7f 3.5 1.1
K- 7f +7f - 7f + 2.7 0.9
K- 7f +7f 12 6 119
K07fo < 6
D+ -0 +
-> K 7f 1.5 0.6 104
K- 7f +7f + 3.5 0.9
The channels listed in table 5.2

account only for the lesser part of
the total width. The unidentified
decays are not found because of
110', 1.78cM{KTtH.92 GeV Ecm =3.mGeV
Reccil ill)iinslO' neutral particles, small branching
fractions and low detection
efficiencies.
Of particular interest is the

decay D+ + ROTI+ which should be
suppressed if sextet enhancement holds
(see 5.132). The experiment 98 ) yields
14 16 IS 20 22 21 26
Recoil Mass To K! It-' r(D++RoTI+) / r(Do+K-TI+) = O.70O.23
and there seems to be little suppression.
Fig. 5.6a Withi n errors the two channels DO + K- TI+
and D+ + ROrr+ are consistent with the
relation (5.16).
The other branching ratios have been compared 98 ) to the statistical

mode1 61 ). The experimental values come out too small except for the K-TI+TIo
where B is too high.
312 H. SCHOPPER
Finally it should be mentioned that the average charge multiplicities

could be determined 98 ) as
for 2.3 0.2

2.3 0.3
which is in fair agreement with the statistical model.
A large amount of data concerning hadronic decays of D and D* were

taken at SPEAR at 4.03 and 4.4 GeV where the total cross section shows
peaks. From these data the results for the D mesons are corroborated, and
the D~ mesons could be established.
The r1ARK 1 detector ~Ias used for

this experiment, and again the iden-
tification of kaons by time-of-flight
.. measurements is essential. The decay
12 0
K111f Ktlt;
\
modes of the DO and D* to various
0
~l ~
final states 101 ) is shown in
0 fig. 5.7. The results for a . Bare
summarized in table 5.3. As one sees
Ks n:' It" KsTt'TC"
there is not much energy dependence
~ ~1:
~o
~ and the results at 3.77 GeV are con-
~ 20
co
..... firmed. In particular the "exotic"
ia 0 Kt. It; It It- state K-~+~+ is seen again, whereas
~
z
~ 160 the Cabibbo forbidden decays are

absent (fig. 5.8).
80
~ Higher multiplicities like K3~, K~~~
o 1
are favoured over lower multiplicities
~~
K'!t" It'
80 like K~. The K3~ state is dominated
by Kp~ while the contribution from
LO
II K*~~ is small. If the maxima at 4.03
o 1.8 2.0 1.6 1.8 2.0 and 4.4 GeV in the total cross section
1.6
EFF [[TlVE MASS (StV/e 1)
are attributed entirely to charmed
Fig. 5.7 meson production, then the identified
Fig. 5.8 Search for DO, D+ Cabibbo forbidden

decay modes at 4.028 GeV
Table 5.3: o B in nb for various D decay modes atthree

values of Ec . m.
Mode E
c.m. (GeV)
3.774 4.0?R 4.414

D K+1I 0.27 0.05 0.57 0.11 0.30 0.09
K01l+1I- + c.c 0.44 0.11 1.09 0.30 0.91 0.34
K+1I1I+1I- 0.34 0.09 0.830.27 0.91 0.39
+-
11 11 -- < 0.04 --
K+K- -- < 0.04 --
Total D 1.05 0.15
observed modes 2.49 0.42 2.12 0.53
0+ K01l+ + c.c. 0.150.05 < 0.18 --

K+/1I 0.34 0.05 0.40 0.10 0.33 0.12
+-
11 11 11 -- 0.03 --
314 H.SCHOPPER
hadronic decays add up to only about 10 % of all decays. The reasons for
this small percentage are undetected neutrals and low detection effi-
ciencies. As will be shown below at these energies the D mesons are pre-
dominantly produced via or associated with D* states.
From the D decays it could also be inferred that parity is violated as

should be expected for a weak decay. The DO ~ Krr decay mode is a natural
spin-parity state (JP = 0+, 1- , ... ) whereas D ~ K rrrr is compatatible with
unnatural spin parity assignment, since 3 pseudoscalars cannot be in a JP 0+
state. The cases JP = 1- and 2+ can be ruled out by studying the population
of the Dalitz plot I20 ).
If charm is a conserved quantum number, D and D* mesons can only be

produced in association. The most likely channels are for DO mesons
e+e- ~ DO DO
DO 0*0 + Do D*o (5.25)
D*o 0*0
and similarly for D+.

A possible way to find D* is to determine the recoil mass
M~ecoil = (Ecm - IpL + M2)2_ p2 against a DO with mo~en~um p and mass M.
This has been done by the SLAC-LBL group using Mark 1 0, 9) .The recoil mass
distribution against a DO is shown in fig. 5.9. Narrow peaks are seen at
Mrecoil -- 1860, 2005 and 2145 MeV/c 2 and a broader peak at 2440 MeV/c 2.
The first two peaks can be interpreted by the first two channels of (5.25).
The third peak could be due to the third channel but e+e- ~ DO D**o could
compete. However, this interpretation is ruled out, as can be seen from
two upper parts of fig. 5.9. The curves are calculated on the basis of
e+e- ~ D*o D*o and one sees a shift and broadening of the peak at 2145 MeV/c 2
at Ecm = 4.02 GeV to about 2200 MeV/c 2 at 4.41 GeV, exactly as expected for
this reaction.
The broad peak at 2440 MeV/c 2 could be due to multibody processes such
as O*O*rr or to production of charm states with higher mass.
The corresponding recoil mass

distribution against a 0+ in the exotic
+ +
channel K n-n- has a much more severe
background. Nevertheless a peak at
2010 MeV/c 2 was observed (fig. 5.10)
which can be interpreted as e+e-~ 0+0*-
and charge conjugate. Indications for
(0)
":..
...... _0 the other peaks might also be pre-
Ec... '4.028 GeV
>
II sent.
..,~
Ii 20 R~tDi I iI ins! D'
0.
i;
.IJ
E
Z
::>
0 ... ) 0
~
:0:
1800 1900 2000 2100 2200

~-~~-~---T"""
(b)
_0
Ec ,", ' 4 4 15 GeV
........
u
>
~
Fig. 5.10
o
<"
Ii The 0* and 0*+ being established
0.
... 100
i II
Ec .. '3 9 -46
as peaks in the recoil mass distri-
::> bution against 0 and 0+ we shall now
Z
turn to the decays of these particles.
Obviously the decays O*O~ OOy and
0*+ ~ O+y should be possible since
they just involve a spin flip of the
c quark (transition between HFS
I~ 2000 2500 states). If the mass difference is
Reco il man (MeV I e') large enough, the strong interaction
decays 0* ~ DOnO, O+n- and
Fig. 5.9 Recoil spectra against D 0*+ ~ O+no, OOn+ should also occur.
316 H. SCHOPPER
These decays are difficult to detect at e+e- energies below 5 GeV since
in this case the pions originating from the 0* have low momenta and with ex-
perimental momentum cut-offs at about 100 MeV/c the detection efficiencies
are low. For this reason measurements in the range E = 5.0 to 7.8 GeV were
performed 121 ). The invariant mass distribution of K ~ shows a clear 0 peak.
If the K combinations in 00 peak are combined with the TIt one expects a 0*
peak in the invariant mass plot. The quantity that is precisely measured is
the mass difference M(Oon) - M(Oo). The distribution of this difference is
shown 89 ) in fig. 5.11. From these data the mass difference 0*+ - 0 is found
to be 145.3 0.5 MeV/c 2 or equivalently
the Q value for the decay 0*+ -+Oon+
is only 5.7 0.5 MeV. Because of this
(al
small phase space the electromagnetic
20 decays 0* -+ Oy can compete with the
strong decays. The mass spectrum of
5 to 7.8 GeV
fig. 5.11 gives also an upper limit on
I, the wi dth of the OM wi th r( OM) < 2 tleV.
For a precise mass determination

it is of advantage to go to a e+e-
energy just above threshold since in
this case the momentum errors have less
(bl
10 influence on the ~ass values (see 5.211).
In fig. 5.12 the momentum spectrum of
0 mesons decaying into 2 and 3 par-
ticles taken at 4.03 GeV is shown.
140 14!> I!>O I!>!> 160 16!> In fig. 5.12a the various processes
mo" - "'0 (MeV..t;2)
contributing to the 0 spectrum are
shown. The peak at about 200 MeV/c is
Fig. 5.11
due to 0*0*. Three channels contribute:
0* -+ OOno. 0* -+ OOy and 0*+ -+ OOn+.
The distribution for the n decays should
beGaussian(curve A. B). whereas for y decays the distribution is dN/dp - P which
results in a triangular shape (curve C). The 0* -+ OOno decay is clearly seen
and also the evidence for 0* -+ OOy is there. For the 0*+ -+ OOn+ only a faint
indication might be recognized. A precise value of the 0* mass can be inferred
from the measured momentum.
The peak at 500 MeV/c and its

shoulder is due to DD* and charge con-
jugation: The central value determines
EGII 4.020 CoV
M(D) + M(D*) while the shape of the
_nt"", spectra
shoulder and peak determine the rala- 20~~"~"'~1~~~~"~""~
10)
tive contributions of the different
channel s.
Finally the peak at about 750

MeV/c is due to direct DODo pro-
duction. A similar, but simpler,
situation holds for the D+ spectrum
(fig. 5.12 c). 20
....
......
>-
A fit 70 ,89,98,lo4) to all these '"
:::E
data gives the results shown in ~
table 5.4 (isospin constraints have :u

c..
10
been used in the fit). In addition '"c:

values for the production processes ...,'"
:0-
have been obtained from these data.

Since they have already been 1~
discussed in chapter 4.41, they will not be

considered here further. 10
Finally it should be mentioned

that a detailed study122) of angular
distributions has shown that the
spins of D* are compatible with 1 as
expected. A: 0 ..... 0"- [,0: 0' 0
000 -~O
The approximate equality of the B: 0' 00' 0

F: 0' 000
0 0 00
decay probabil it i es of DlfO -+ Y DO and
H:
(Gaussianl
Dlf + -+ n+D o is in qualitative agree-
ment with the theoretical expecta- d
(ap - p)
tion 85 ). The main reason is, of course,
the suppression of the strong decay Fig . 5. 12
by its small phase space.
318 H. SCHOPPER
Table 5.4: Masses and decays of D*
Particle Mass (MeV/c)2

0*0 2006. 1.6
0* 2008.6 1.0
1863 3
1874 5
Decay Branchi ng rati

0.75 0.05
0.60 0.15
Mass differences (MeV) Q-values (MeV)
0+ - D 5.0 0.8 0*0 .. D0 7.7 1.7

0*+ - 0* 2.6 1.8 0* .. 0+ - - 1.9 1.7
(0+ -D) - (0*+_0+) 2.4 2.4 0*+ .. D + 5.7 0.5
0*+ .. 0+ 5.3 0.9
The masses of D and 0+ agree well with those determined at 3.77 GeV
(see chap. 5.211).
In fig. 5.11b the nonexotic KTr7T combinations are shown. From the
data at the position of the D* mass the limit 89 )
< 16 % (90 % Cl) (5.26)
can be derived which puts a limit on DO - DO mixing (see chap. 5.14). From
the number of events where the K in the DO recoil has the same sign as
in DO a different limit on DO - DO mixing can be obtained 89 ):
< 18 % (90 % Cl) (5.27)

If we introduce the parameter
t: =- - - - - - - - - (5.28)
where N(K) gives the number of events with kaons of opposite and same sign
respectively. one obtains experimentally t: ~ 0.8.
For complete DO - Do mixing t: = 0 and for no mixing t: = 1. The experiments
are thus compatible with the non-existence of a neutral flavour changing
current.
The data also yield the limits for the Cabibbo forbidden decays
(5.29)
r(D o ~ K+K-) < 0.07

r(D o ->- K1T)
and hence tg2e < 0.07.
The observation of semileptonic decays of charmed mesons is very inter-

esting since the hadronic current appears only once and therefore a rather
clean test of the weak charm current is possible (see chap. 5.12).
Semileptonic decays of charmed mesons with an electron in the final

state were observed for the first time by the DASP collaboration 124 ) in
spring 1976 at the time when the hadronic decays were found at SPEAR. Very
soon PLUT0 125 ) could show that the electrons are associated to KO mesons as
one would expect for the decay D ->- e v R (see fig. 5.13 and 5.14).
Semileptonic decays into electrons require a clean identification of the

electron. In the DASP experiment this was achieved by Cerenkov counters. in
the PLUTO experiment by the showering property of electrons. Following a
whim. nature has chosen the mass of the D meson (1.86 GeV/c 2) very close to
the mass of the heavy leptontll) (1.80 GeV/c 2 ). Hence the correspondii~g lepton
320 H. SCHOPPER
Semileplonic Decays e+e-- TtTt+X .... DESY (PLUIO]
LI..--'-~---'-----<J <D ~~-fl---fl~/ (!)

I I S I / I /
M....t pon~, HluIG,V HI. oJ GfY HtmtlGf'I
l; "M ASS PEAK
41-1~';-
5 ----!-I---!,/ a> I I I / l:-~I-;I~I~/ (!)
M.ss at poon PI" HIKoJ G,Y HI IG,V MI IG,y
ASSOCI Ale 0 ElECTRON SIGNAl
Fig. 5 . 13
,e"-r.lo other DESY

(PLUIOI
460<l4n n <SU H,
f'e- -~":"'" D* D
I..Kev
\0
Ko ntlOuum
40
ECftlIifV "
Fig . 5.14

--"
e'p " - pt.~2prongs
v 3.0
>
399 < Eell < S.20 Gpy
Cl MISlDE NTlF IEO HAO"ONS
1- 2.0
- " - HE AV~ lEPTONS
";ii
N
[
. ..
A'
.... OS 10 tS
bl~
'0'0 p. (G pY/c)
Fig . 5 .1 5
spectra cover similar energy regions and one has to find ways to separate
the two phenomena. This can be done on the basis of two criteria:
I.} The charmed mesons tend to produce final states with more than 2 charged
tracks, w~ereas the heavy lepton has predominantly two tracks in the
final state.
2.} The D mesons decay mainly into K'mesons whereas the heavy lepton has
a very low branching ratio for kaons.
In this way one can distinguish two types of events: high multiplicity,
large kaon content versus low multiplicity, few kaons. These two kinds of
events have been invetigated 127 } quite carefully and it could be shown
that the meson and lepton decays can be separated quite well. The average
number of charged kaons was found to be 0.90 0.18 per multiprong event
and 0.07 0.06 per 2-prong event. This also shows that for charmed par-
ticles (i.e. high multiplicity) the weak current couples strongly to strange-
ness in accordance with the GIM mechanism.
The inclusive electron spectrum measured by DASp126) for the range

Ecm =3.99 to 5.2 GeV is shown in fig. 5.15 before having been corrected. One
notices that the contamination from the heavy lepton is quite small. In fig.
5.16 the electron spectrum taken close to E ~4 GeV is shown after subtracting
128) cm
the background. Two theoretical curves are drawn for the decays D-> K e \)
and D-> K* e\). The accuracy is not good enough to di sti ngui sh between the two
channels. The best fit is obtained for a mixture of both. Since the contribution
of the two decay modes to the D decays depends sensitively on the V, A structure
of the weak current, it seems very important to improve the measurements such
that a separation of the two channels becomes possible. This might be achieved
e+e-- e:!: +~2prongs
40 399 < EeN < 4.08 GeY
---- 0 -toil<
30 - O-.vK-(892)
20
10
Pe(GeV/c)
Fig. 5.16
322 H. SCHOPPER
by the DELCO detector at SPEAR which covers a much larger solid angle with
C-counters than DASP (60% compared to 7% of 4~). So far the statistical accuracy
that could be obtained does not yet allow such a separation 129 ). Some results
with lower statistical accuracy have recently also be obtained by MARK 1 130)
All these data are in good agreement.
DASP data 126 ) taken at higher Ecm show no drastic change of the spectrum
which is not trivial since they were taken above the thresholds for F and
charmed baryon production.
From the spectrum displayed in fig. 5.16 one can infer a limit on the
leptonic decay D-r ev which would produce a peak near Pe ::: 1 GeV/c. One finds
a(D-r eVe) / a (D-r eX) < 0.09 (90 % Cl) in agreement with the theoretical
expectation (see chap. 5.11). The Cabibbo-suppressed decay D -r ~ e v (see
chap. 5.12) produces a spectrum similar to those from D -r K e v and D -r K* e v
but it is shifted to higher momenta. From the fits to the measured spectrum
it can be excluded that the decay D -r ~ e v is the sole leptonic decay mode.
Integrating over the spectra and comparing to the total cross section
one can derive the leptonic branching ratios. These have already been dis-
cussed in chap. 4.42.
Recently an indication of the semileptonic decay of the F meson

has been found by DASP 123) . The n's found at Ecm ~ 4.16 GeV are associated
to electrons (fig. 4.1~ hatched events). A value for the branching ratio
cannot yet be given, since the evaluation of the efficiencies is still
under progress.
El ectron spectra for the decays F -r n e v, n' e v, e v have been

calculated 131 ) .
No semileptonic decays of charmed mesons vlith muons in the final

states have been seen. This is because the detection and identification
of muons with momenta below 1 GeV/c is very difficult.
5.23 Discovery of F meson~
As has been explained in chap. 5.13 the lowest states of the F mesons
tend to decay to final states containing ss which results in nand n'. Since
it is more difficult to detect n's than K's the discovery of the F mesons has to
be considered much more difficult that those of the D mesons which are charac-
terized by kaons in the final state. A first indication of F production was ob-
tained from the inclusive yield of n which shows an abundance at Ecm = 4.4 CeV
(see chap. 4.43). Since n is a frequent byproduct of n' decay, a search for
n includes a search for n' .
The DASP group105) succeeded in detecting F mesons by their decay

F->-nn; however, this could be achieved only by using a "trick". The signature
of this decay channel is not specific enough to separate the F decays from the
large background. The theoretical prediction was that the first excited state
F* (the ortho cs state having parallel spins) should be close to the spin-zero
ground state F (see chap. 2.22) and the favoured decay should be F* ->- F y
with a photon energy of about 100 MeV. This low-energy photon delivers an addi-
tonal selection criterion. What has been observed are events of the type
e+e - ->- F* F4 nn
(5.26)
ly
~yy
F
low
A good event is characterized by 3 photons, one of which has a low energy, the
other two forming a mass in the n region. In addition an identified charged pion
with momentum above 0.6 GeV/c had to be seen. These events could, however, also
originate from the reaction
e+e - ->-
F (5.27)
4n n
Lyy
324 H. SCHOPPER
e+e- -FF~ FyF

L~rr
4.36 ,; W s 4,48 GeV
[}ASP PRELIMINARY
25
>
'""
M
+
2D
~
~ +
~ 1.5
.,t
+
1.0 15 20 2.5
M~~ (GeV)
Fi g. 5.17
e'e-- FF~ FyF

l'!Jn:!
06 :S W :S 4.48 GeV
DASP PRELIMI ARY

8
o~~ __~~______~~~
1.0 1.5 2.0
M~~ (GeV)
X2 < 7 IM~~ - M~sl< 250 MeV
Fig. 5.18
The observed events were fit for both processes. The most recent results are
shown 123 ) in fig. 5.17, where the invariant 7Tn mass is plotted against the
recoil mass. A clear clustering of events can be seen corresponding to (5.26)
or (5.27). In fig. 5.18 the projected n7T mass distribution is shown. The peak
in the mass bin 2.025 to 2.050 GeV is attributed to the F.The events at lower
masses could originate from A2 decays.
The conclusion on the masses is
m{F) 2.03 0.06 GeV/c 2

(5.28)
m{F*) 2.14 0.06 GeV/c 2
where the uncertainty originates essentially from the ambiguity (5.26) or

(5.27). The mass difference F* - F can be determined directly from the energy
of Yl ow and the result is
m{F*) - m{F) = 120 40 MeV (5. 29)
The interpretation of the HFS splitting has already been discussed in chap. 2.2.
It agrees astonishingly well with the theoretical expectation.
These data were taken at Ecm = 4.4 GeV where the inclusive n production
shows a prominent peak (see fig. 4.13). However, it would be very interesting
to observe FF production just above its threshold at 4.06 GeV (see table 4.2)
where FF* is not yet possible, in complete analogy with DO production at
3.77 GeV. The right place for such a search seems to be the peak at 4.1 GeV
and indeed a clear n peak has recently been found 123 ) in the yy invariant mass
distribution (fig. 4.15). In this case no low-energy photon was required, of
course. The shaded events in fig. 4.15 show events with an additional electron.
One notices that the n peak is correlated with electrons indicating that a weak
decay \~as involved. No n peak was seen for e+e- energies 4.0 to 4.06 GeV, i.e.
below FF threshold.
326 H. SCHOPPER
Mass r - - - - -21~0:
-- 50
-
MeV F*+tt
yI""'" "'V
2100
2000 n
F+ t.
y \ 2030~50
142.7 MeV \ Y 140.2: 0.9 MeV
!1} n+
1900 \\
DO t+ D+ h
18 63.3 lB5B.~
. 0.9 :!:. 0.9
Fig. 5.19
Perhaps the most striking success of the charm model is the discovery
of the charmed mesons predicted by the model.The most important results are:
1.) Both the isospin doublet DO, D+ and the isospin singlet with strange-
ness F+ have been seen. As expected the lowest states are the ones
with J = 0 (quark spins opposite). The corresponding states with
J = 1 have also been found and the HFS splitting agrees with the ex-
pectation of the potential model (see chap. 2.22). These ortho states
decay to the ground states by strong or electromagnetic transitions.
The transition rates present no problems to their theoretical inter-
pretation.
The mass spectrum of the charm mesons and the observed transitions
are summarized in fig. 5.19.
Higher excited states, e.g. P states, have not yet been seen.
2.) The weak decays of the ground states of charm mesons are compatible with
the expectations of the minimal model of weak interaction (see chap. 5.1).
The rule lIC = liS = 1 which is impl ied by the minimal theory is drama-
tically confirmed by the dominance of kaons in the final states. Also
the Cabibbo structure of the charm current is evident by the suppression
of lIC = 1, 1I S = 0 decays. Parity violation has been demonstrated.
3. ) The chi ra 1ity structure of the charm current coul d not been verifi ed
yet. The accuracy of the semileptonic decay spectra is not sufficient
to distinguish between V-A and V+A. A distinction between D ~ e v K and
D ~ e v K* would help, since the first decay goes by pure V, whereas
the second contains V and A components. Polarization
measurements would be the most direct way to clarify the
chirality structure. Ao polarization from charmed baryon decays could
offer a possible way.
4.) No sextuplet enhancement is found in charm meson decays. As a conse-
quence, the 1I1 = 1/2 enhancement found in ordinary strange particle
decays cannot be explained by 20-plet enhancement.
5.) No DO - 50 mixing is found. The present accuracy excludes the existence
328 H.SCHOPPER
of neutral currents with ~C =0 of order GF.

6.) Although there is evidence for charm baryons in neutrino reactions,
they could not be identified so far in e+e-. Indirect evidence is pro-
vided by steps in the p and A inclusive cross sections.
So far all the observations on charm mesons are compatible with the minimal
theory. This model, however, cannot be exact if embedded in a world with 6 leptons
and more than 4 quarks. Hence it would be very interesting to find deviations.
6. The Upsilon
6.1 Experimental results
In 1977 a Columbia-Fermilab-Stony Brook collaboration l32 ) under the

guidance of L.M. Lederman bombarded nuclear targets with 400 GeV protons and
observed the reaction
p + nucleus ~ ~+ + ~- + anything (6.1)
in a two-arm spectrometer which allowed the identification of the muons and

the measurement of their momenta. From these the invariant mass of the muon
pair can be computed and the distribution of the invariant masses is shown in
fig. 6.1.Superimposed on an exponentially decreasing background one notices
enhancements at the J/~ and ~' masses
and a broad structure between 9 and
dO.,. 10 GeV/c 2 . If the smooth background is
-32 dmdyly:o -0.3~Y~+0.3 subtracted one obtains the data dis-
J/ljI
played in fig. 6.2. The exper;mental re-
-33 solution in this region is 200 MeV.
1jJ'
-34
c::
g -35
<>
:>
c::
.....
~ -36
C>
......
N
~ -37 8 9 10 II
moss (GeV)
, -38 Fig. 6.2
Hence the width of the structure cannot

-39
be due to the experimental resolution
but indicates some complicated structure.
-40~~~~~-+.~~~~~~
2 4 6 8 10 12 14 16 18 Indeed, just looking at the data of
p.+}J-- MASS (G e V I
fig. 6.2 one would assume peaks around
Fig. 6.1 9.4 and 10 GeV/c 2 .
330 H. SCHOPPER
The authors have fitted the data by two different assumptions, by

2 or 3 Gaussian peaks, respectively. The position of the peaks and their
branching ratio x production cross section are given in table 6.1.
The existence of the T" is statistically not certain, of course. Not

much can be said about the width of the peaks. In the fits it is assumed
that their widths are small compared to the experimental resolution of about
200 MeV.
From the figures of table 6.1 one can extract the following data which will
be important for the discussion in chap. 6.2.
Table 6.1: Resonance Fit Parameters
2 peak 3 peak
M1 9.41 0.013 9.40 0.013 GeV
r { cryi

B do 0.18 0.01 0.18 0.01 pb

y=O
r.{
M2 10.06 0.03 10.01 0.04 GeV
B do 0.069 0.006 0.065 0.007 pb

cryi y=O

M3 --- 10.40 0.12 GeV
r"i B dOl
dy y=O
--- 0.011 0.007 pb
x2 / DF 19.3 / 18 14.2 / 16
2 peaks 3 peaks
tl('r') - M(')") 650 30 ~eV 610 40 MeV

:1(')'" ') - M(Y) 1000 120 t-ieV
Ratio of '(I /y 0.38 0.04 0.37 0.04

B do ')'" '/Y 0.06 0.04
dyjy=O
Clearly it seems very important to resolve experimentally the complex

structure in its components. This might be possibly only by producing the T
in e+e- collisions. Although not being designed for such high energies, DORIS
might be pushed to reach the T region and this will be tried in spring 1978.
6.2 Possible interpretation of the T family
After the discovery of the T there appeared a flood of theoretical

papers making the assumption that the T family is just a replay of the
J/~ story. This means that there exists a family of QQ-bound states, where
Q is a heavy quarks of charge 2/3 or 1/3 and the mass of Q is about
half the T mass, i.e. mQ ~ 4.6 GeV.
~~!_~~~i!~!iQ~_~~~r9i~~
Even before the detection of T, the standard model with a linear
confining potential had been extended 133 ) to larger quark masses. The
excitation energies above the lowest-lying state are shown in fig. 6.3
as functions of the quark mass. The
~-,~------------------~ I OOO
essential point is that for a linear
potential the excitation energy is re-
duced proportional to mQ- 1/ 3 Since the
600 continuum boundary on the other hand goes
up one expects more bound states. For
--"!--~400 mQ ~ 4.6 GeV not only the IS and 2S states
are bound but also the 3S level.
200 zoc
mo ( Gtv) Fig. 6.3

332 H. SCHOPPER
According to this model one expects M( T') - M( T) = 420 MeV and

M( T") - M( T) = 750 MeV, values which are smaller than the experimen-
tal ones. Indeed, the experimentally found splittings imply
M( T') - ~1( T) '" M(lji') - M(JN) in contradiction to the prediction of the
linear confining potentia'. This perhaps is not very surprising since the
naive standard model failed to give the proper positions of the charmonium
P states. One possible remedy was to assume that the confining potential
is not a Lorentz vector but a scalar (see chap. 1.232). Indeed it was
shown 134 ) that a scalar linear potential gives splittings independent of mQ
and of the right magnitude.
Another possibility is to look for a vector potential with mQ-inde-

pendent excitation energies. Indeed it could be shown 135 ) that a logarith-
mic potential V(r) - ln r does exactly that. It also gives good account
on the charmonium levels and is even better than the linear potential,
as far as the P levels are concerned. However, it predicts too large E1
transition rates and gives uncomfortably large relativistic corrections 87 ).
If the interpretation of the T family as a new QQ system is right

then the study of the excitation levels of this system together l'lith that
of charmonium offers us a unique possibility to learn something about the
quark-quark forces. The spectrum of the QQ system could even be much
richer than the charmonium decay scheme 133 ) but we shall not discuss it
here further, since the scaling of the excitation energies is not clear
yet.
The interesting question is whether T is made of charge 2/3 (top)

or 1/3 (bottom) quarks. A number of authors have tried to answer this
question by comparing the measured production cross section to different
models 136 ). Clearly B da/dy is 4 times larger for a 2/3 charge than for 1/3.
0
Most of the theoretical estimates are in favour of 1/3 (for a comparison see
Ref. 132), but the uncertainties are so large that no definite conclusion can
be drawn.
A surprising fact is the large ratio of B . do/dy for T' compared to T.

In analogy to ~' one would expect 87 ) a large branching ratio for the transi-
tion T' + T + 2TI. For the ~'/~ couple the ratio is R = (1.7 0.5) % whereas
for T' / T one has R ~ 0.4. This could be taken as evidence against the assumption
that all states belong to the same family. However, from a somewhat naive appli-
cation of ACD one can find 8?) plausible arguments that the T' + y + 2TI tran-
sition is suppressed by OZI mechanism by a factor of about 10. More experimen-
tal information is needed to clarify this issue.
Of particular interest will be the hadronic decays of the T family.

The masses of these particles are large enough that hadron jets originating
from gluons should be well developed. This might provide an excellent test
of the ideas of QCD. Because of conservation laws a 3S1 state can decay
into 3 photons (e.g. ortho positronium) or 3 gluons (J/~ family, see
chap. 3.2) but not 2 photons or gluons. A 3p state on the other hand decays
into 2 photons or 2 gluons. Hence it is expected that T (9.4) decays into
3 gluons which should appear 137 ) as 3 coplanar hadron jets.
The excited states of T are expected to decay to P states by electro-

magnetic transitions and one expects 138 ) decays like
T' (10.0) + y + 3P2 (9.8) + y+ 2 g + y + 2 jets
where one finds two jets with distinct lab energies (~ 5 GeV) accompanied
by a monochromatic photon. The angular distribution allows us to differentiate
between jets originating from gluons and from quarks.
If the QQ interpretation of T is right, particles with Qq etc. should

also exist where q is one of the 4 lighter quarks. Obviously they are
analogous to the D and F mesons and their ground states would be stable
with respect to strong interactions. Their weak decays and possible mixing
between BO - BO and TO _ fOhas been discussed(B o = bq, TO = tq)136)137)138)139)
6.3 Sequential quarks and leptons
The existence of the charm quark made it possible to group the 4 quarks
and 4 leptons into weak SU(2)W doublets and a minimal theory with GIM mechanism
could be formulated for the weak interaction. It is compatible with
334 H. SCHOPPER
all experiments (see chap. 5.1), except for the non-observation of parity
violation in atomic transitions 14o ) which, however, needs confirmation
and the small CP violation in K decays. The beauty of this
model is the symmetry between quarks and leptons which is needed
to cancel the triangle anomalies which would otherwise have destroyed
the renormalizability of the weak and electromagnetic gauge theory141).
With the discovery of another heavy lepton T and probably another

quark it is tempting to extend the Weinberg-Salam-~Jard-GIM model by adding
further lefthanded quark and lepton doublets:
c' :, ... J (6.2)

s'
The extension to a 6-quark--6-lepton model was first proposed by Kobayashi

and Maskawa 142 ). It retains the natural suppression of ~S = 1 and ~C = 1
effects as observed and preserves the phenomenological success as far as
neutrino scattering is concerned.
A very important point is, however, that the 6-quark model gives CP
violation in a natural way. This is achieved by a generalization of the
Cabibbo structure of the hadronic current (see chap. 5.1). The most general
way of mixing the lefthanded charge 1/3 quarks is143)
d' = c 1 . d + sl c3 s + sl s3 b
s' =-sl c 2 d + (-s2 s 3e io + c1c 2c3) . s + (c 1c2s 3 + s2c3 e

io
) b
io io ) . b
b' sl s 2 . d - (c 1s 2c3 + c2s 3 e ) . s + (c l s 2c3 + c2c3 e
(6.3)
whereas the 2/3 charge quarks u, c and t remain unchanged. Here the
abbreviated notation sl = sin 01' c l = cos 01' s2 = sin 02' '" has been
used and 01 is identical to the Cabibbo angle. The 4-quark model is repro-
duced by 02 = 03 = 0 = O.
Since the original Cabibbo structure agrees well with observations

c 3 has to be close to one. A more quantitative analysis (Ellis et al. 136 ))
of the coupling in lJ decay and in (ud) and (us) gives s1 2 s3 2 < 0.003 and
with s1 2 '" 0.05 this means s3 2 <0.06. Hence 83 cannot be much larger than
the Cabibbo angle 81 but it may, of course, be much' smaller. The restrictions
on 82 are much weaker and depend on the quark masses. For mt '" 5 GeV one
finds s2 2 < 0.2 and for mt ~ mW '" 65 GeV s2 2 < 0.03.
The phase parameter a introduces CP violation and appears only if one

has at least 6 quarks. From the measured CP violation parameter ~ 10- 3
one deduces that a cannot be arbitrarily small: Isinol > 3xlO- 3 for
mt = 5 GeV and Isinol > 6x 10- 4 for mt = 65 GeV.
Because of the limits for 82 and 83 one might speculate if they are
exactly zero. This has quite amusing consequences. The decay of the strange
particles to ordinary hadrons is only possible because of the d-s mixing.
Without this "feed-through" the lightest strange particles would be abso-
lutely stable. With 6 2 = 6 3 = 0, b would not mix with d and s, as can
be seen from (6.3) and as a consequence particles like bq would be stable
if the b quark is lighter than the t quark or v.v. Such stable particles
of a new kind of matter could be captured in nuclei and the consequences
for the ensuing X-ray spectra and formation of hypernuclei have been recently
discussed in detail I44 ).
If the interpretation of the T family as a QQ spectrum of a new kind

of quark turns out to be right then a number of questions arises, e.g.
1) is there one more quark to complete the couple (t,b)?
2) which one is the lighter, t or b?
3) is the structure of the weak current such as expected
from a minimal theory?
4) are 6 quarks and 6 leptons the basic elements of matter
or are there more sequential leptons and quarks?
5) are electrons and muons elementary contituents or are
they made up of smaller entities having 1/3 and 2/3 charges?
It is hoped that the e+e- storage rings PETRA, CESR and PEP which will
come into operation soon might make it possible to shed some light on
these exciting and fundamental questions.
336 H. SCHOPPER
Appendix 1
Table Al: Decay modes of the $ into mesons
Topology Decay Mode Observed Number Efficiency Branching Ratio

of Events
KSMM < 4 0.23 < 8.9 10- 5

KSKL
IT +IT - a11 1.7 1.7- 0.040 (1.6 1.6) x 10- 4
K+K- a11 1.8 1.4 - 0.036 (2.0 1.6) x 10- 4
KSK IT'' a11 126 15 0.13 (2.6 0.7) x 10- 3

K<>'K "'0 + ROK- o 45 7.8 0.044 (2.7 0.6) x 10- 3
K+K"'- + K-K- + 48 7.7 0.040 (3.2 0.6) x 10- 3
KORiIIiIIO + ROKiIIiIIO 1 2.7 0.007 < 2.0 x 10- 3
K+K.... - + K-K.... + -1 2.7 0.006 < 1.5 x 10- 3
IT +IT -K+K- a11 205 17 0.076 (7.2 2.3) x 10- 3

K-oRuO+ R*oK-- o 40 8.4 0.016 (6.7 2.6) x 10- 3
~
K-oR*o 1.5 4 0.048 < 0.5 x 10- 3
K.... oR....o 2.5 4.5 0.009 < 2.9 x 10- 3
4>lT +IT - 23 5 0.043 (1.4 0.6) x 10- 3

oI>f ~ 1 0.023 < 3.7 x 10- 4
+-+-+- (3.1 1.3) x 10- 3
IT IT IT IT K K a11 30 6 0.026
+- + -
tn 1T 1T TT ::; 3 0.013 ~ 1.5 x 10-
3
K+K-K+K- all 19 5 0.075 (0.7 0.3) x 10- 3

4>K+K- 14 5 0.040 (0.9 0.4) x 10- 3
oI>f' 6 3 0.020 (0.80.5) x 10- 3
IT+lT-K+K- lTo all 309 50 0.073 (1.2 0.3) x 10- 2
wK+K- 22 12 0.068 (0.8 0.5) x 10- 3
wf' - 2 2.4 0.034 < 1.6 x 10- 4
4>n 5 2.5 0.013 (1.0 0.6) x 10- 3
IT+lT-K+K- MM 4>n' ~ 2 0.011 < 1.3 x 10- 3
2(lT +IT -)lTo a11 675 40 0.17 (4.0 1.0) %

WlT +IT - 348 40 0.14 (6.8 l,g) x 10- 3
wf 81 20 0.11 (1.g 0.8) x 10- 3
Ao + A 36
P 2 P 2 12 0.018 (8.4 4.5) x 10- 3
3(lT+lT-)lTo a11 181 26 0.062 (2.9 0.7) %
w4lT 140 30 0.044 (8.5 3.4) x 10- 3
4(lT+lT-)lTo a11
13 0.014 (9.0 3.0) x 10- 3
4
IT +IT- IT (polTo + PlT") 153 13 * 0.12 (1.l 0.3) ;:
2(lT+lT-) a11 9 ..
76 0.19 (4.0 1.0) x 10- 3
3(lT +IT-) a11
32 7 0.80 (4.0 2.0) 10- 3
Modes marked with an asterisk were calculated from a sma11er

da ta sample.
References
1. H. Fritzsch, M. Gell-~lann and H. Leutwyler, Phys. Lett. B47 (1973),365

D.J. Gross, F. Wilczek, Phys. Rev. D8 (1973),3633
S. Weinberg, Phys. Rev. Lett. ~(1973),494
2. T. Appelquist and H.D. Politzer, Phys. Rev. Lett. 34(1975),43
T. Appelquist, A. De Rujula and H.D. Politzer, Phys. Rev. Lett.34
(1975),365
A. De Rujula and S.L. Glashow, Phys. Rev. Lett. 34(1975),46
3. A. De Rujula, H. Georgi, and S.L. Glashow, Phys. Rev. ~(1975),147
R. Barbieri, R. Kogerler, Z. Kunszt, and R. Gatto, Nucl. Phys. B105
(1976) ,125
3.a A. De Rujula, R.C. Giles and R.L. Jaffe, MIT-preprint CTP No. 635,
June 1977
4. J.D. Jackson, Summer Institute of Particle Physics, SLAC, August, 1976
and papers quoted there
5. E. Eichten, K. Gottfried, T. Kinoshita, J. Kogut, K.D. Lane,
T.-M. Van, Phys. Rev. Lett. 34(1975),369
6. M. Krammer and H. Krasemann, Nuov. Cim. 32A(1976),394
7. H.J. Schnitzer, Phys. Rev. Lett. ~(1975),1540 and Phys. Rev. D13
(1975),74
8. J. Pumplin, W. Repko and A. Sato, Phys. Rev.Lett.~(1975),1538
8.a A. Martin, Phys. Lett 67B(1977),330

H. Grosse, Phys. Lett.68B(1977),343
H. Grosse, and A. Martin, Nucl. Phys. B132(1978),125
9. C.E. Carlson and F. Gross, College of Willam and Mary Preprint
WM-PP-21 (March 1977)
10. A.B. Henriques, B.K. Kellett and R.G. and R.G. Moorhouse, Phys. Lett.
64B(1976),85
11. H.J. Schnitzer, Phys. Lett. 65B(1976),239
12. A. Arneodo and G.L. Kane, Preprint, Michigan Univ., UM-HE 77-5(1977)
13. T. De Grand, R.L. Jaffe, K. Johnson and J. Kiskis, Phys. Rev. D12(1975),
2060
338 H.SCHOPPER
14. E. Eichten, K. Gottfried, T. Kinoshita, K.D. Lane and T.-M. Van,

Phys. Rev. Lett.36(1976),500
15. S. Okubo, V.S. Mathur and S. Borchardt, Phys. Rev. Lett. 34(1975),236
Ngee-Pong Chang and C.A. Nelson, Phys. Rev. D35(1975),1492
16. A. Kazi, G. Kramer, D.H. Schiller, DESY 75/11 (1975) and
Acta Phys. Austriaca 45(1976),65
S. Borchardt, V.S. Mathur, S. Okubo, Phys. Rev.Lett.34(1975),38
B.J. Edwards and A.N. Kamal, Phys. Rev.D39(1977)66
17. G. Kallen, Elementary Particle Physics, Addison-Wesley, 1964, pp29
18. H.J. Schnitzer, preprint, Brandeis Univ., March 1977
19. W. Celmaster, Harvard Univ., Preprint 1977, HUTP77-A024
20. H. Fritzsch, preprint CERN, July 1977
21. C.E. Carlson and F. Gross, preprint, College of William and Mary,
Williamsburg, WM-PP-21, March 1977
22. A.B. Henriques, B.K. Kellet and R.G. Moorhouse, Phys. Lett. 64B(1976),85
23. A. De Rujula, H. Georgi and S.L. Glashow, Phy. Rev. Lett. 37(1976), 398
see also K. Lane and E. Eichten, Phys. Rev.Lett. lZ(1976),477
24. K.D. Lane, and S. Weinberg, Phys.Rev.Lett.37(1976),717
25. J.J. Aubert et al., Phys.Rev.Lett. ~(1974),1404
26. J.-E. Augustin et al., Phys.Rev.Lett.33(1974),1406

27. G.S. Abrams et al., Phys.Rev.Lett. 34(1975),1181
W. Tannenbaum et al., Phys. Rev. Lett.36(1976),402
28. G.J. Feldman, Summer Inst. on Particle Physics, SLAC, August 1976,
SLAC-PUB 1852
recent data see G. Goldhaber, European Conf. on Particle Physics,
Budapest, July 1977, LBL-6732
29. B.H. Wiik and G. Wolf, Les Houches Summer School 1976, DESY Report
77/01 (1977)
30. A.M. Boyarski et al., Phys. Rev. Lett. 34(1975),1357
V. LUth, SLAC-PUB-1599 (1975)
31. W. Braunchweig et al., Phys. Lett 63B(1976),115

32. L. Criegee et al., DESY Report 75/32 (1975)
J. Burmester et al., 1976 Tblisi Conference and DESY Report 76/53 1976)
33. C. Bemporad, 1975 Stanford Conference, p.113
34. W.Braunchweig et al., Phys. Lett.57B(1975),297
and Phys. Lett.63B(1976),487
35. R. Barbieri, R. Gatto and R. Kogerler, Phys. Lett.60B(1976),183
36. S. Okubo, Phys. Lett. 5(1963),105
G. Zweig, CERN Rep. TH-401,412 (1964)
J. Iizuka, K. Okada and o. Shito, Prog.Theor. Phys. 35(1966),1061
37. J. Burmester et al., DESY report 77/50 (1977)
and Phys. Lett. 72B(1977),135
38. F. Vanucci et al., Phys.Rev. D15(1977),1814
39. J.S. Whitaker et al., Tblisi Conference 1976
40. W.Braunchweig et al., Phys. Lett.67B(1977)243
41. W. Bartel et al., DESY rep. 76/40(1976), Phys. Lett.64B(1976),483
and Phys. Lett. 66B(1976),489
42. W. Bartel et al., DESY report 77/70 (1977)
43. G. Knies, DESY report 77/74 (197n and Proc. of the 1977 International
Symposium on Lepton and Photon Interactions at High Energies,
Hamburg, August 1977, p.93,
G. Alexander et al., DESY Rep. 77/72 (1977)
44. A. De Rujula, H. Georgi, and S.L. Glashow, Phys.Rev.D12(1975),154
and Phys. Rev. Lett.34(1975),46
45. H. Fritzsch and P. Minkowski, Nuov. Cim. 30A (1975),393
46. H. Fritzsch and J.D. Jackson, CERN report TH-2264 (1976)
47. W. Braunchweig et al., Phys. Lett.67B(1977)249
48. B.J. Edwards and A.N Kamal, Phys.Rev.Lett.39(1977),66
49. T.F. Walsh, DESY report 76/13 (1976) and unpublished information
50. R. Brandelik et al., DESY report 78/01 (1978)
51. M. Krammer, DESY report 78/06 (1978)
340 H.SCHOPPER
52. W. Braunchweig et al., Phys. Lett.67B(1977),243

53. W.Bartel et al., DESY report 77/70 (1977)
54. C.J. Biddick et al., Phys.Rev.Lett.38 (1977), 1324
55. J.S. Whitaker et al., Phys. Rev. Lett. 1I(1976) , 1596
V. Blobel, XII Rencontre de Moriond (1977)
56. W.D. Apel et al., Phys. Lett. 72B(1978),500
57. W. Braunchweig et al, Phys. Lett. 67B(1977)249
58. W. Tannenbaum et al., Phys. Rev. Lett. ~(1975),1323
59. J.D. Jackson, European Conf. on Particle Physics, Budapest, 1977

60. G. Feinberg and J. Sucher, Phys.Rev. Lett. ~(1975),1740
61. C. Quigg and J.L. Rosner, FNAL-report 77/40 (1977)

62. F. Wilczek and A. Zee, Phys. Rev. Lett. 40(1978),83
63. H. Krasemann and M. Krammer, DESY report 77/34 (1977) and
Phys. Lett. 70B(1977)457
64. D.H. Boal, Phys. Rev. Lett. 37(1976), 1333
65. M. Greco and M. Krammer, DESY report 77/30 (1977) and Phys. Lett.
69B(1977),313
66. W. Braunschweig et al., Phys. Lett. 57B(1975),407
B. Wiik, Tblisi Conference 1976
67. D. Badtke et al., Tblisi Conference 1976
68. W. Bartel et al., Tblisi Conference 1976
J. Olsson, Proc. of the 1977 Intern. Symposium on Lepton and Photon
Interactions at High Energies, Hamburg, 1977, pp 117
69. U. Timm, DESY report 77/52 (1977) and European Conference on Particle
Physics, Budapest, 1977
70. G. Goldhaber, European Conference on Particle Physics, Budapest 1977
and LBL-6732 (1977)
71. M.S. Chanowitz and F.J. Gilman, SLAC-PUB 1746 (1976)
72. A.B. Henriques, B.H. Kellett and R.G. Moorhouse, Phys. Lett. 648 (1976),85
73. J.D. Jackson, Phys. Rev. Lett.37(1976),1107
74. R. Barbieri, R. Gatto and E. Remiddi, Phys. Lett. 61B(1976),465

75. J.D. Jackson, European Conference on Partilce Physics, Budapest, 1977
76. T.W. Appelquist and H. Georgi, Phys. Rev. D8(1973),4000
A. Zee, Phys. Rev. D8(1973), 4038
77. A.M. Boyarski et al., Phys. Rev. Lett.34(1975),764
78. U. Timm, DESY report 77/52 (1977) and European Conference on Particle
Physics, Budapest, 1977
79. G. Knies, DESY report 77/74 (1977) and Proc. of the 1977 Intern.
Symposium on Lepton and Photon Interactions at High Energies,
Hamburg, 1977, pp 93
J. Burmester et al., Phys. Lett. 66B(1977)395
80. S. Yamada, Proc. of the 1977 Intern. Symposium on Lepton and Photon
Interactions at High Energies, Hamburg, 1977, pp.69
81. R. Brandelik et al., Phys.Lett. 73B(1978), 109
82. G.J. Feldman, SLAC-PUB-2000 (1977)
83. A. Backer, PHD-Thesis, University Siegen, DESY report F33-77/03 (1977)
84. P.A. Rapidis et al., Phys. Rev. Lett. 39(1977), 526
85. K.D. Lane and E. Eichten, Phys. Rev. Lett. ll(1976) , 477
86. J. Kirkby, SLAC-PUB-2040 (1977) and Proc. of the 1977 Intern. Symposium
on Lepton and Photon Interactions at High Energies, Hamburg, 1977,
pp 3
87. K. Gottfried, Proc. of the 1977 Intern. Symposium on Lepton and Photon
Interactions at High Energies, Hamburg, 1977, pp 667
88. G. Goldhaber et al., Phys. Lett. 69B (1977), 503
89. H. Sadrozinski, Proc. of the 1977 Intern. Symposium on Lepton
and Photon Interactions at High Energies, Hamburg, 1977, pp.47
90. J.G. Korner, M. Kuroda, G. Schierholz, Phys. Lett. 70B(1977), 106
90.a J.G. Korner and M. Kuroda, Phys. Rev. ~(1977),2165
91. L.B. Okun and M.B. Voloschin, JETP 23(1976), 369

A. DeRujula, H. Georgi and S.L. Glashow, Phys. Rev. Lett. 38(1976),317
342 H. SCHOPPER
92. J. Burmester et al., DESY report 77/17 (1977)

93. T.F. Walsh, DESY report 76/13 (1976)
R. Kogeler, G. Schierholz and G. Kramer, Phys. Lett. 65B(1976),441
94. M. Tonutti, Proc. of the 1977 Inter. Symposium on Lepton and Photon
Interactions at High Energies, Hamburg, 1977, pp135
95. J. Burmester et al., DESY report 77/14 (1977)and Phys. Lett. 67B(1977)367
96. R. Brandelik et al., Phys. Lett. 67B (1977), 363
97. V. LUth et al., Phys. Lett. 70B (1977), 120
98. A. Barbaro-Galtieri, Proc. of the 1977 Intern. Symposium on Lepton
and Photon Interactions at High Energies, Hamburg, 1977, p. 21
99. R. Brandelik et al., DESY report 77/44 (1977)
100. G. Wolf, Intern. School of Phy~ics "Enrico Fermi", Varenna, July, 1977
101. M. Piccolo et al., Phys. Rev. Lett. 39(1977) 1503
102. B. Knapp et al., Phys. Rev. Lett. 1Z(1976), 882
103. E.G. Gazzoli et al., Phys. Rev. Lett. 34(1975),1125
104. I. Peruzzi et al., Phys. Rev. Lett. 39 (1977), 1301
105. R. Brandelik et al., Phys. Lett. 70B (1977), 132
106. J.D. Bjorken and S.L. Glashow, Phys. Lett. 11 (1964), 255
107. S.L. Glashow, J. Illiopoulos, and L. Maiani, Phys. Rev. 02 (1970), 1285
108. H. Fritzsch, Intern. School of Physics "Enrico Fermi", Varenna, July
1977, CERN-TH 2359 and Intern. Summer Institute for Theor. Physics,
Bielefeld, September 1977
109. S.L. Glashow, Nucl. Phys. ~ (1961), 579
110. A. Salam and J.C. Ward, Phys. Lett. 1l (1964), 168
111. S.Weinberg, Phys. Rev. Lett.~ (1967), 1264
112. D. Fakirov and B. Stech, Heidelberg THEP 77-8 (1977) and contribution
to the 1977 Intern. Symposium von Lepton and Photon Interactions at
High Energies, Hamburg, 1977
113. T.F. Walsh, DESY report 77/76 (1977) and Proc. of the 1977 Intern.
Symposium on Lepton and Photon Interactions at High Energies, Hamburg,
1977, pp 711
114. A. Ali and T.C. Yang, Phys. Lett. 65B(1976), 275

and DESY report 77/54 (1977)
114.a J. Ellis, M.K. Gaillard, and D.V. Nanopoulos, Nucl. Phys. B100 (1975),313
115. R.L. Kingsley et al., Phys. Rev. Q!l(1975), 1919
116. M. K. Gaillard, B.W. Lee, J.L. Rosner, Rev. Mod. Phys. 47(1975),277
117. S.L.Glashow and S. Weinberg, Phys. Rev. D15 (1977),1958
E.A. Paschos, Phys. Rev. D15 (1977), 1966
118. A.M. Boyarski et al., Phys. Rev. Lett. 35 (1975), 196
119. D.L. Scharre et al., SLAC-PUB 2019, September 1977,
Phys. Rev. Lett 40(1978), 74
120. J.E. Wiss et al., Phys. Rev. Lett. 37(1976), 1531
121. G. Feldman et al., Phys. Rev. Lett. 38(1977), 1313
122. H.K. Nguyen et al., Phys. Rev. Lett.39 (1977), 262
123. DASP-Collaboration; DPG-Spring Meeting, Heidelberg, March 1978,

and G. Wo11, private communication
124. W. Braunchweig et al., DESY report 76/37 (1976) and Phys. Lett. 63B(1976),417
125. J. Burmester et al., DESY report 76/50 (1976), and Phys. Lett. 64B(1976),369
126. R. Brandelik et al., Phys. Lett. 70B(1977), 387
127. R. Brandelik et al., Phys. Lett. 70B(1977),125
128. A. Ali and T.C. Yang, Phys. Lett. 65B (1976), 275
129. J. Kirkby, SLAC-PUB 2040, November 1977
130. J.M. Feller et al., Phys. Rev. Lett. 40(1977),274
131. A. Ali and T.C. Yang, DESY report 77/54 (1977)
132. S.W. Herb et al., Phys.Rev. Lett.39(1977),252
L.M. Lederman, Proc. of the 1977 Intern. Symposium on Lepton and
Photon Interactions at High Energies, Hamburg, 1977, pp 567
133. E. Eichten and K. Gottfried, Phys. Lett. 66B(1977),286
134. D. Pignon and C.A. Piketty, preprint Nov. 1977
Lab. de Phys. Theor. de 1 'Ecole Norm. Sup.
135. C. Quigg and J.L. Rosner, FERMILAB Pub-77/82 (1977)
344 H. SCHOPPER
136. C.E. Carlson and R. Suaya, Phys. Rev. Lett. 39(1977),908

R.N. Cahn and S.D. Ellis, Univ. Michigan Report 76-45 (1976)
J. Ellis et a1., Nuc1. Phys. B131(1977), 285
T. Hagiwara, Y. Kazama and E. Takasugi, Phys. Rev. Lett. 40(1978), 76
137. K. Koller and T.F. Walsh, DESY report 77/68 (1977)

138. M. Krammer and H. Krasemann, Phys. Lett. 73B(1978), 58
139. A. Ali, CERN report TH.2411 (1977)
A. Ali and Z.Z. Aydin, DESY report 78/11 (1978)
140. P.G.H. Sandars, Proc. of the 1977 Intern. Symposium on Lepton and
Photon Interactions at High Energies, Hamburg, 1977, pp 599
141. C. Bouchiat, J. I11iopou10s and Ph. Meyer, Phys. Lett.38B(1972),519
142. M. Kobayashi and K. Maskara, Prog. Theor. Phys. 49(1973), 652
143. H.Fritzsch, CERN TH-2539 (1977)
144. R.N. Cahn, Phys. Rev. Lett.40(1978), 80
DIS C U S S ION
CHAIRMAN: Prof. H.F. Schopper

Scientific Secretaries: J.H. Field, T. Uematsu, S.L. Wu
DISCUSSION No.1
SIZZARRI:
My question is on the F*-F mass difference. You
said it is determined mainly by the ~ energy which is
given as 120 40 MeV, but you have a cut at 140 MeV with
a resolution of 60 MeV. Since this cut presumably excl-
udes a large number of events I wonder if the observed
peak at 120 MeV could not be artificially produced.
SCHOPPER:
The energies of the ~ rays are determined not by
direct measurement, but by kinematical (2c) fits to the
events. The energy resolution obtained in the fit is
much narrower than given by the direct measurement. The
large spread of 40 MeV quoted in the ~ ray energy comes
not from the errors on the individual fits, but from the
uncertainty in the production hypothesis, i.e. whether
it is:
345
346 DISCUSSION
SMITH:
)'c
What fraction of the F F -+ F y F events are lost by
the cut on '( ray energy?
SCHOPPER:
I don't have the number. Wiik should be able to

answer it when he comes; but it is certainly a large
fraction of the signal.
WU:
In the SPEAR result there is a peak in R at about
4.43 GeV. I understand that your FF* events come from
a scan from 4.32 to 4.44 GeV. Why do you choose to run
on the left side of the peak? What exact beam energies
do these four FF* events correspond to?
SCHOPPER:
The absolute scales of beam energies at SPEAR and

at DORIS are slightly different, and the peaks in R
show in slightly different places.
WU :
If we look at the rich structure in R between 3.9

to 4.6 GeV, how much is understood from heavy leptons,
charmed mesons D and F? How much room is left for new
discoveries in this region?
SCHOPPER:
I shall go into detail in my third lecture, but

there is not much room left for new discoveries.
DEO:
Since you are g~v~ng an explanation of the new

particle spectrum in terms of potentials, can you tell
us about model independent results. Maybe Martin will
be able to tell us about this.
DISCUSSION 347
MARTIN:
It shall be covered in my lectures.
CELMASTER:
I have a comment to make about potential models and
various other things you said this morning and I also
have a question afterwards.
There is one thing that can be observed from all
these models. By and large, most details about charm-
onium were postdicted and not predicted. Correct pred-
ictions are common to most of these models, where these
models differ are in the "postdictions", to my view this
is ugly and I don't think that it is very valuable to
overly refine these models. My bet is that they will
have minor failures for "bottomonium". Another thing to
consider is the following:
All predictions are based on the Born approximation.
In particular this holds for the "Coulomb-perturbation"
prediction 3'P. _JR
& t -=.8
"'?1 -~"P
0
If you solve the Schrodinger equation directly for those
three states, that is, by including the spin orbit etc.
forces in the potential, it turns out that
3 ~"'D
1; - r1 =.5
3'P. _ l'P,
1 D
Perhaps this shouldn't be taken too seriously, as other
types of perturbative effects might compete with this.
However, we do see that it is possible, within the exist-
ing framework to explain such small effects; it is not
necessary to introduce long range scalar-scalar inter-
actions etc. Finally, regarding ~c; I think that it is
not wise to try too hard to explain its mass. There are,
as Professor Lipkin has said, two other difficulties
which no one can explain in the potential model context,
that is, its very small hadronic width and the low mag-
netic transition from J/~ smaller by a factor of 15 from
prediction. This leads me to believe that X (2.8) is
not understandable in potential models, unless it is ~
the 'tIc
348 DISCUSSION
CELMASTER:
My question is: can you repeat why you use for the
F signature the'~data in the way described during your
lecture?
SCHOPPER:
The preferred decay of the F is into S8. We do not

use the K's as a signature because the D's also decay
into K's. Whereas, the F decays into,~ much more freq-
uently than the D. In addition the soft y from the
f''c -. F decay was required to give an even cleaner signature
LIPKIN:
Is the spin dependence of the potential treated only

in perturbation theory, or is the Schroedinger equation
solved in the spin dependent potential? The splitting
between the ~/'" and the tt (2.8) is 300 MeV and the split-
ting between the ~/* and t~e f'is 600 MeV, so the spin
splitting is not small compared to orbital splittings
and perturbation theory is not obviously justified.
SCHOPPER:
No, the Schroedinger equation was solved in the new

potential and new wave functions are found.
CARTER:
You said this morning that the problem oftS split-

ting among the P-states was patched up by introducing a
Lorentz-scalar exchange piece into the potential. I do
not understand why a Lorentz-scalar exchange force gives
a spin-orbit force in the non-relativistic limit.
MARTIN:
The point is that the non-relativistic contributions

of the spin-dependent terms in the potential, (giving
rise to spin-orbit and tensor splittings), have different
forms for the scalar and vector interaction, (see
J.D. Jackson's recent talk at Budapest), and so by adjust-
ing the relative size of these terms the level spacing
DISCUSSION 349
can be changed.
OLSEN:
The way I understood the raw~J spectra that you
showed this morning, was that they corresponded to events
which had already passed the ~~ , (one less than 140 MeV)
2 charged particles. Is that correct?
SCHOPPER:
Yes.
OLSEN:
These raw distributions seem to show a signal to

noise of about 1 to 2, or 1 to 3; on the other hand in
the final scatter plot of the F signal, the signal to
noise is much improved, i.e. 2 to 1. Do you under-
stand how this happens?
SCHOPPER:
The improvement comes from the requirement of a
charged it in one of the spectrometer arms.
OLSEN:
What is the momentum acceptance of these spectro-
meter arms?
SCHOPPER:
Please reserve the question for Bjorn Wiik.
OLSEN:
,
Are all the n. $ observed consistent with having
originated from F's?
350 DISCUSSION
SCHOPPER:
Yes, when there is the additional requirement of a

low energy, 140 MeV) ,(-ray.
LIPKIN:
I should like to make the following comment on

potential model calculations of the F-P': splitting. You
should be careful not to exagerate the importance of any
prediction of the F-F* splitting, because it fits into
the simple systematics described by my "Crazy Mass Form-
ula": M~,,-Mr= M~- M~ to a very good approximation.
So any model which agrees with this formula for ordinary
vector and pseudoscalar mesons will also probable agree
for the F and F*.
DISCUSSION No.2 (Scientific Secretaries: K. Wacker,

T.J. Weiler, S.L. Wu, J.H. Field)
CABIBBO:
You presented data on J /~ decay into 1.I)1llt and <p1tX

which indicate that singly forbidden processes are not,
if we subtract resonant contributions, dominant over dou-
bly forbidden ones. I would like to remark that subtract-
ing resonant contributions is against the spirit of the
Zweig rule. According to the standard folklore, allowed
processes are dominated by resonant contributions in each
channel. If I subtract all resonant contributions, I
should remain with no allowed contribution. One could
perhaps see the resonant character of allowed contribut-
ions as the reason for its dominance over OZI forbidden
contributions.
SCHOPPER:
But if there are no resonant contributions, isn't

it reasonable to ask how the singly and doubly connected
diagrams compare?
DISCUSSION 351
CABIBBO:
This experimental result is not contrary to what

you would expect, but it could mean that the process is
completely dominated by resonance production.
PREPARATA:
The value (:r-+ cpTtlt /"J-'f C01111 ') ~ ~ which you showed this
morning to remain after subtraction of known resonances
is in good agreement with the following idea. The 1r~
system is in S-wave and for the 0+ I=O resonances the
strange - nonstrange quark admixture is, as experiment-
ally suggested, non-negligible.
LIPKIN:
This would suggest the comparison of '\>'Jt"1t to oo"''lt

non-resonant states is really cheating here because there
is a resonant contribution in Cp1tn which has not been
taken out. The 'Jt1[ spectrum seems to be almost entirely
0+ resonant states.
I have another remark. The way to look for hadronic
decays of the ~ is to note that the decay goes via
co -+qq and subs~quent decay of a qq state with the quan-
tum numbers of the~. The hadronic channels strongly
coupled to the 'yt would be the places to look, e. g. <t> <t>
and ltAz . pp would be a strong decay but the background
is just impossible. The pp mode is expected to be small
because the ~ is known to have a very weak coupling to
the nucleon.
What is the experimental situation on cp~ and ltA2.
decay modes?
SCHOPPER:
People have looked for pp because it is experiment-

ally simple to do so. One can identify the p by time of
flight. To look for~~, which decay into Kls, is more
difficult. One has just learned how to look for Kls, and
one can start looking for ~~ in the future with a large
solid angle detector.
352 DISCUSSION
WEEKS:
Again the Zweig rule: You presented cases where the

interpretation of Zweig suppression was the weak coup-
ling of hard gluons to quarks. Do you personally accept
this as a universal mechanism for Zweig's rule, and what
is the consensus of opinion on this matter?
LIPKIN:
For the explanation of Zweig's rule presented by

Schopper a particular set of diagrams are selected, namely
those with qq annihilation into gluons. The point is
there are other diagrams which have nothing to do with
the use of perturbation theory. For example, you have
the diagram k p
~1(
Note that there are no gluon intermediate states.
SCHOPFER:
How about J/~ decay?
LIPKIN:
If you believe the KK intermediate state is respon-

sible for <f~ f>1t, you_would expect those J /-f states
which are above the DD threshold to decay through the
DD intermediate state. You would have, for example
A \,/11
'i' ~
-
PD -+ c..u -+ plt '
SCHOPPER:
Below the threshold you need the gluon diagrams.
LIPKIN:
So there is a danger in comparing the decay of the

J/~ (below DD threshold) wi th the decays of the <{' and
f' (above KK threshold). For the latter there are
extra diagrams.
DISCUSSION 353
WETZEL:
I would like to point out that by counting gluons

one obtains a very poor estimate for the fJ(1514)_~~
rate. According to the Rosenfeld table this process has
not been seen. A~ upper bound has been given using f,f'
interference in KK production, which is of the size
~2 (f'~ 1t.~) / (,'te
f'-. lei.) .{ .1 '1... This is much smaller than
the corresponding value 9r'2 (c.p .... flt )/lJ."Z(\f""Ic.k.> for the <p
in spite of the fact that the f' decay can proceed via
two gluons while <f decay require s three gluons.
LIPKIN:
The problem of the fl decays has been discussed at

length in the Erice lectures of 1976. It is pointed out
there that the f" -'> /tx decay can be understood in terms
of the diagram
The purists say that you should not expect asymptotic

freedom to work at all for light quark systems.
DISCUSSION No.3 (Scientific Secretaries: J.H. Field,

K. Wacker, T. Weiler, S.L. Wu)
CABIBBO:
You mentioned this morning that our group was pred-

icting a branching ratio of 1.4% for (D-,>~""X)/(D~~\I).
This was indeed our prediction in 1974, based on the
hypothesis that whatever mechanism was active in the
enhancement of 61=1/2 decays would also be active, with
the same strength, in charm decay. In successive papers,
which appeared in 1975, and were presented, e.g., at the
Palermo conference, we pointed out that a detailed comp-
utation based on QCD led to a smaller enhancement, i.e.
to a branching ratio for ~~X/all of the order of 6%.
This number should therefore be the prediction of QCD.
We should keep in mind that a similar computation
of the enhancement of AI=1/2 in nonleptonic decays of
strange particles fails to give a complete account of the
354 DISCUSSION
enhancement observed there. We think that the extra

enhancement in that case should be attributed to low
energy considerations, such as those proposed by
Preparata.
PREPARATA:
All of the enhancement can be traced to "long dis-

tance" (low energy plus Regge) effects, as S. Nussinov
and I have shown, a long time ago, for the baryon non-
leptonic decays.
LIPKIN:
I have a comment on the OZI rule violations via

diagrams of the type:
~,t'
1L
These diagrams can be calculated for on-shell KK states,
and the results agree with experiment. Tne corresponding
proces2 J 1,1.
't' -
-
DO - O,U),u:l, ... ~ plt, '71.,... is not possible
I"
wi th DD on shell so the <{>+ flt results cannot be extra-

polated to T/,\>-+pions. When one tries to describe these
processes in terms of quark diagrams, a twisted diagram
results, viz.
\1=:."+
Topologically, the two loops can be pulled apart to give

a disconnected diagram, so the process is forbidden. But
strictly speaking the suppression is only effective at
asymptotically high energies relative to the thresholds
of the intermediate states.
A question results from these diagrammatic consider-
ations. Is_ '1"','1>111) ... -+ DiS ~ (J),Ib' (J)", ... --+ pions important
abov~ the DD threshold? Then ~ ("V l"4pions) should jump
at DD threshold. Similarly, ~ln -+ (KK+pions) should jump
at PF threshold.
DISCUSSION 355
PREPARATA:
If we want the OZI rule to be good for the J/~ decay
the only way to keep the offending diagram
1)
~ D
from destroying it is to assume that ~l~~DD)~lJ'I'\l~t>'5)
i.e. a very big violation of SU(4). This is borne out
by calculation in the Geometrodynamical approach as I
have shown last year at the Trieste Topical Conference
on Particle Physics.
LIPKIN:
The contribution from on shell intermediate states

cannot be cancelled by off shell contributions. We know
all the states which can contribute on shell, so nobody
can invent a new cancellation mechanism with new states.
But off shell contributions involve all the higher re-
sonances that have never been seen and model builders
are free to put in anything they please to cancel it.
ROLANDI:
o
In reference to the recoil mass spectrum from D ,
I cannot understand why the peak "on the right" from
where you reconstruct the recoiling mass of (i5!no ) is

so narrow?
SCHOPPER:
The reaction takes place near to the threshold so
that the recoil momenta involved are all very low, as is
the Q-value of the rl! -. bltC) decay. At higher energies
the peak is significantly wider due to the smearing
effect you are thinking of.
RECENT RESULTS FROM DASP
B.H. wiik
Deutsches Elektronen-Synchrotron DESY
Hamburg, Germany
In this talk I will discuss recent data obtained with the

DASP detector at DORIS. A comparison of these data with other ex-
periments and a discussion of the theory can be found in the lec-
tures by H. Schopper and A. Martin at this School.
Since early this year we have collected data at c.m. ener-

gies between 4.0 GeV and 5.2 GeV for an integrated luminosity of
-1
nearly 6000 nb These data have so far yielded results on the
following topics:
1) Inclusive spectra of charged kaons.
2) Evidence for the F meson.
3) The origin of inclusive electron events.
4) Semileptonic decays of charmed mesons.
5) Leptonic decay of the T.
6) Semihadronic decays of the T.
An exploded view of the detector 1 ) is shown in Fig. 1. The

detector is made of two identical magnetic spectrometers positioned
357
358 B. H. WIIK
Fe 5 OK SR
Fig. 1 Exploded view of the DASP detector
symmetrically with respect to the beam axis. A non-magnetic inner

detector is mounted in the field free region between the arms. A
charged particle emitted at the interaction point in the direction
of one of the magnet arms, traverses a scintillation counter So
close to the beam pipe, two proportional chambers, a threshold
Cerenkov counter, a second scintillation counter SM' and a wire
spark chamber, before reaching the magnet. The Cerenkov counters,
which are filled with Freon 114 at atmospheric pressure (n = 1.0014),
detect electrons with momenta above 10 MeV/c and pions above 2.8
GeV/c. Beyond the magnet the particle passes through five wire
spark chambers, a wall of time-of-flight counters, and shower
counters. The shower counters are followed by iron planes inter-
spersed with wire chambers at depths of 40 cm and 80 cm and a
RECENT RESULTS FROM DASP 359
plane of scintillation counters at a depth of 70 cm of iron. The

spectrometer identifies particles by means of the range informa-
tion, the pulse-height in the shower counters, information from
the time-of-flight counters and the Cerenkov counters. Electrons
are identified over the whole momentum range by a signal from the
Cerenkov counters. This identification is verified by time-of-
flight information for momenta less than 0.35 GeV/c and by pulse-
height in the shower counters for momenta above 0.35 GeV/c. Muons
are identified by range for momenta above 0.7 GeV/c. The time-of-
flight information separates pions from kaons up to 1.6 GeV/c and
protons from pions and kaons up to 3.0 GeV/c. The geometrical
solid angle covered by the detector is 0.9 sr.
A particle emitted towards the inner detector traverses the

following elements: one of 22 scintillation counters surrounding
the beam pipe, two proportional chambers (in part of the accept-
ance), four modules - each made of a scintillation counter hodo-
scope, a 5 rom thick lead converter and a tube chamber with two or
three planes of proportional tubes, and finally a lead scintilla-
tion counter hodoscope 7 radiation lengths thick. The detector
covers about 9 sr and is particularly well suited for photon de-
tection down to energies of 50 MeV. The direction and energy of
a photon are measured with an accuracy of 25 mrad and ~E/E =
0.14/IE(GeV), respectively.
The detector was triggered for all the data reported by a

single charged particle which traversed the magnet and gave sig-
nals in two scintillation counters (SO and SM) in front, and a
time-of-flight counter and a shower counter in the rear. For most
of the data, the magnet current was set so that charged particles
with momenta of 0.1 GeV/c, or above, were accepted.
360 B. H. WIIK
1. INCLUSIVE SPECTRA OF CHARGED KAONS

2 3)
It is generally believed that the step observed' in ,the
total e+e- annihilation cross-section near 4 GeV in c.m. is due
to the production of the charmed mesons DO, D+ and F+ made of the
quark combinations (c~), (cd) and (c;).
The charmed 4 ) quark predominantly decays into a strange quark.

Therefore, the charmed mesons will preferentially decay as fol-

lows: DD + (K- ), D+ + (K ) and F++(n, n', ~, K+-
K ). The step
in the total cross-section is thus expected to be reflected in a
similar increase in kaon production.
The total cross-section for e+e- + KX as measured 5 ) by the

DASP group is plotted in Fig. 2 versus c.m. energy for kaon momen-
ta between 0.3 GeV/c and 1.6 GeV/c and also for all momenta. The
The cross-section increases by a factor 2.5 for c.m. energies be-
tween 3.6 GeV and 4.03 GeV and then seems to level off at higher
energies in general agreement with the total annihilation cross-
section as expected. A simple kinematical observation can now be
used to show that these kaons indeed result from charm production.
At threshold the charmed mesons are produced at rest, hence kaons
resulting from their decays must have energies less than one half
of the beam energy. This is shown in Fig. 3 where the invariant
cross-section (E/4n p2) do/dp is plotted as a function of kaon
energy for 3.6 GeV and 4.05 GeV. Figure 3 shows clearly that the
step in the total kaon cross-section is caused by kaons with ener-
gies below E/2 as expected if they are the decay products of a
heavy particle produced nearly at rest.
We now estimate the fraction of the total cross-section as-

. . .
soc~ated w~th charged kaon product~on.
.
F~gure
4a sows
h r)
Rnew( vs
Rtot (/;) _ Rtot (3.6) = Rtot(/S) _ 2.5 as measured 3 ) by the PLUTO
group. Also plotted is ~ew(/S) = ~(rs) - ~(3.6). So far, we
have tacitly assumed that a single mechanism is responsible for
the increase in the total cross-section near 4 GeV. As discussed
(J" (nb)
15r-------,r------,-----,
10
?0.3<p<1SGeVIc
+aamomenta
O~--~I~--~~--~
3 4 5 6
Vs(GeV)
Fig. 2 The cross-section for the sum of K+ and K produc-

tion versus c.m. energy for momenta between 0.3
and 1.6 GeV/c (open circles) and for all momenta
(black circles).
362 B. H. WIIK
10 _ I
I
~ vslGeV)
l-
t- o 3.6
-
x 4.05
N
~
<!)
r:-
-
:0
c:
blo-
'0'0
t-
t- t
wf} t-
0.1 ::-
1j j t \ 1
1
0.01 t
.L-....1----L_L---1-_.l.--.I....1----L.---l_..I....-......I.---I
0.5 1.0 1.5
E I GeV)
Fig. 3 The invariant cross-sections E/(4rrp2) dcr/dp for
the sum of K+ and K- pro duct~on
. at c.m. ener-
gies of 3.6 GeV and 4.05 GeV.
I
3 -
(a)
2
R"ew
~~~ ~++ +4 +
~ f~+
+-
1 r-
<} <}
-
o .6.
1T
-1 RKew
-2 -
I I
4 Rnew
1
-
3 - -
(b)
~-
r-
~~ ~
lr- ~~ \ . +*++H + -
9 .:.
or-~A
1- +R~ew -
~ R"ew,no H.L.
2r- I I
I
3.5 4.0 4.5 5.0
Vs (GeV)
Fig. 4 Comparison between the total cross-section for new

hadron production and the cross-section for (K;+K-)
both in units of the point cross-section.
a) Rnew = R - 2.5.
b) Rnew corrected for the contribution of a heavy
lepton of mass 1.95 GeV.
364 B. H. WIIK
below, rather convincing evidence now exists that a heavy lepton

+ -
with mass around 1.9 GeV is produced in e e annihilation and that
this lepton decays into final states with a small ratio of kaons
to .
p~ons. Hence the heavy 1epton contr-~' b '
ut~on to Rnew( vs
~s) s h ou ld
be subtracted in order to obtain the charm contribution. The re-

suIt is shown in Fig. 4a. These data show that on the average every
charm event contains a charged kaon.
Results in agreement with these data have been obtained for neu-
6) 7)
tral kaons by the PLUTO group and the SLAC-LRL group
2. EVIDENCE FOR THE F MESON
The charm model predicts that there should exist mesons with
+ -
both charm and strangeness. The lowest member of this set F (cs)
should decay predominantly into an ss system leading to final states
containing KK, . n or n'. A step in the n yield would. therefore.
be a signature for pair production of F mesons. Note that an in-
clusive n search also includes a search for n' since n' + nrrrr oc-
curs with a branching ratio of 68%.
, e + e - annihilation near D* thresh-
Experience has shown that ~n
-* and D*-*
old DD D production is copious. In analogy to this we as-
-* and F*-*
sume FF F production to be enhanced. The first excited
state F * is ,
pred~cted to be close to the F such that F* + yF or
F+ + rrDF (electromagnetic only. strong forbidden by isospin) are
the favoured decay channels. To enhance the signal we searched 8 )
for final states with an n and a low-energy photon. We identify
the n by its decay into two photons.
For the multihadron events we selected events with two charged

tracks and more than two, but less than six, photons. At least
two photons had to have energies above 0.1 GeV. The vector sum of
the momenta of these photons had to be above 0.3 GeV/c but less
than 1.2 GeV/c. All such pairs of photons are used to reconstruct
the m mass. The events were divided into two classes depending
yy
on whether the event contained a photon with energy less than
0.14 GeV or not. The myy distribution of events containing a
low energy photon is plotted in Fig. 5 for all energies except
4.4 GeV. and in Fig. 6 for 4.4 GeV. The solid line represents a
smooth estimate of the uncorrelated background obtained by forming
the invariant mass of photons from different events having the
same number of photons. The background spectrum is normalized to
the data by requiring all combinations with m > 0.7 GeV to be
yy
ascribed to background. There is a clear n signal at 4.4 GeV, not
observed at other energies. To ascertain that the n is correlated
with a low-energy photon, we also evaluate the m distribution
yy
for events without a low-energy photon and plot the ratio of the
m distribution for events without a low-energy photon. This ra-
yy
tio is plotted in Fig. 7 as a function of myy At 4.4 GeV there
is an enhancement at the mass of the n showing that the n is in-
deed correlated with a low-energy photon.
The occurrence of n production in association with low-energy

photons at 4.4 GeV strongly suggests the presence of F meson pro-
duction near 4.4 GeV via FF* or F*r*. We searched for the two-body
+ +
decay of the F- into nn- by selecting events with the following
topology:
e+e- ~ TC'!: + '1 (l) + YlDlAr + OA~tlU ....S
The pion was measured and identified in one of the spectrometer

arms and it was required to have a momentum above 0.6 GeV/c. The
2
mass of the yy system had to be between 0.35 and 0.65 GeV/c , and
the low-energy photon had to be below 0.2 GeV. A total of 35 can-
didates satisfied these criteria. The events were fitted to the
reaction
366 B. H. WIIK
4.0 GeV
60
40
4.15 GeV
>~
-
~
o
LIl
I II
c
.2 4.5 GeV
g 60
:0
~ 40
-...
u
o
~
.c
E
J
Z
0.25 0.5 0.75 1.0
Fig. 5 Distribution of myy for events with a low-energy

photon 0.14 GeV) for various values of the
c.m. energy. The solid lines are estimates of
uncorrelated photon background.
4,4 GeV
80
>CI)
~
0
Lf')
........
Ul
C
0
.-
0
40
C
..0
E
0
u
20
0.25 0.5 0.75 1.0 1.25
m (yy) in GeV
Fig. 6 Same as for Fig. 5 for IS = 4.4 GeV

368 B. H. WIIK
~} l.! III
.4
.3
.2
f II
.1
~.1!!! III
.1.
.3
.2 I If
-9.4 ~
.1
I f I I II
;>-
0
c: .3
.......
I II
-
~
~.2
.1
4.4 GeV I
- .3 .4
0
0
Ifj!!IlllJ
0:::
.2
4.5 GeV
.1
.4
.3
.2
I I Iff f r I I J
5.1 GeV
.1
0.2 0.4 0.6 0.8
M (yy) in GeV
Fig. 7 Ratio of the number of combinations for events
having a low-energy photon to the number of com-
binations for events not having a low-energy
photon as a function of m
yy
(1)
(2)
These are 2C fits because of the mass constraint m and the re-
yy
quirement that for (1) TIn and the missing vector must have the
+
same mass ~; for (2) the Yl F- system and the missing vector
ow
must have the same mass mF*. A total of 15 events satisfied the
criteria with a X2 of less than 8. The same events also fit the
hypothesis (2).
In Fig. 8 the results of a fit to hypothesis (1) are shown.

Events with a X2 < 8 are plotted in a scatter plot of the recoil
mass minus the (nTI) mass versus the fitted nTI mass. At 4.4 GeV
there is a cluster of events with a TIn mass around 2.04 GeV and a
recoil mass around about 100 MeV higher. A fit to hypothesis (2)
leads to similar results. An average over the values obtained by
the two fits leads to the following values for the mass:
~ (2.03 0.06) GeV

~* (2.14 0.06) GeV
From the energy distribution of. the low-energy photon alone we

find mF* - mF = (0.12 0.04) GeV.
370 B. H.WIIK
- 600 -400 -200 0 200 400 600

M(meas.)-M(Fft.) (MeV/c 2 )
0.30 1..1. GeV

Q25
0.20
..
~ 0.15
u
">GI 0.10
--
C> 0.05 ~
-
t:!
F 0.30
1..0-5.2 GeV
~
0.25 ~
excluding 1..1. GeV .
I
~
~
0.20
0.15

0.10

0.05
0
1.0 1.2 1.1. 1.6 1.8 2.0 2.2
M (lilt) Fitted (GeV/c 2)
Fig. 8 Results of a fit the hypothesis e+e- + FF* with

F* + yF and F + nn as described in the text. At
IS = 4.4 GeV there is a signal at M(nn) = 2.03
GeV with the recoil mass at about 2.15 GeV.
3. THE ORIGIN OF INCLUSIVE ELECTRON EVENTS
Single electrons observed in e+e- collisions are evidence for

pair production of weakly decaying particles. Hadrons with new
flavours or new leptons are two examples of such particles.
We have investigated 9 ) the single electron spectrum for c.rn.

energies between 3.6 GeV and 5.2 GeV for a total integrated lumi-
-1
nosity of 6300 nb Electrons were defined off-line as particles
which gave a signal in the proper Cerenkov counter and had B= 1
(p < 0.35 GeV/c) or gave a large pulse-height (p > 0.35 GeV/c) in
the shower counter. A pion had a probability of less than 10- 3
to pass these criteria. Electron pairs from Dalitz decay or pair
conversion were rejected by pUlse-height cuts on the scintillation
counters mounted before the magnet.
To reduce the background of an electromagnetic origin, the

event must contain at least one non-showering track. This could
be a track in a spectrometer arm identified as a hadron or a muon.
A charged track traversing the inner detector could also be called
non-showering if it fired less than 1.5 tubes per layer, averaged
over all layers which had at least one tube set. Events contain-
ing only two charged particles are particularly sensitive to elec-
tromagnetic background. Here we required less than 1.25 tubes per
plane be activated. A test of these criteria, using well-defined
pions, showed that 95% of the pions but ,fewer than 5% of the
electrons satisfied the tight criteria.
The background to this sample from beam gas events, Compton

scattering on the material mounted in the front of the Cerenkov
counter or from two photon processes was estimated and found to be
small. The total background was (11.5 3.5)% for the two-prong
sample and (15 5)% for the multiprong sample. These values are
in agreement with measurements done at the ~' resonance and at 3.6
GeV.
372 B. H. W"K
After all cuts 60 two-prong and 182 multiprong events with

an electron momentum above 0.2 GeV/c remained. The measured
charged multiplicity distributions (including the electron) for
all inclusive electron events and for those with only charged
tracks are plotted in Fig. 9. The distribution peaks for nch =2
but it is rather wide with events up to a charged mUltiplicity of
eleven.
Observed Prong Distribution in e + e -- e + X
60
Any No. of Y' s
wm No. Y
-
III
c
~
ell
40
Z
20
2 4 6 8 10 12
N prong with e
Fig. 9 The number of inclusive electron events versus

the charged multiplicity observed. The electron
is included. The distribution of events without
photons is hatched.
We have shown 9 ) earlier that the majority of the high multi-

plicity events result from weak decays of a hadron with a new fla-
vour and there is now little doubt that the bulk of these events
are weak decays of charmed hadrons. Indeed, even assuming a heavy
lepton which decays 30% of the time into three or more charged
particles we find that less than 12% of the events with nch ~ 3 can
be explained as heavy lepton production.
We now investigate if charm production can be the source of

events with two charged prongs and no photons.
In the nch 2, ny ~ 0, we observe 37 events. Can these be

explained as leak through from higher multiplicities? The produced
multiplicity distribution has been reconstructed from the known de-
tection efficiencies for charged and neutral particles and the ob-
served multiplicity distribution. We find that only (8.5 2)
events can result from the production of higher multiplicities.
+
Can charm directly produce (e- + non-showering track)? 1) Associated
charmed hadrons are produced, one decaying to e and invisible par-
+
ticles, the other to a non-showering track T- and invisible par-
ticles. 2) Neutral charmed hadrons are produced, one decaying to
+
e and T- plus invisible particles, the other decaying to invisibles.
3) A charged charmed hadron yields the e, a neutral charmed hadron
.
decays invisibly and T 1S produced directly. These production
modes have been considered in detai1 9 ) and we find that they can
account for less than 12% of the two in detai1 9 ) and we find that
they can account for less than 12% of the two-prong photon events.
The charged multiplicity distribution associated with inclusive
electron events cannot be explained by a single mechanism. Whereas
charm production naturally explains the final states of large mul-
tiplicity (n ch ~ 3), it can only account for a small fraction of
the two-prong class. Pair production of heavy leptons is, however,
expected on quite general grounds to populate final states of low
374 B.H.WIIK
~ultiplicity. The multiplicity distribution observed can, there-

fore, in principle, be understood by pair production of a new lep-
ton in addition to charm production. We discuss the properties of
the multiprong sample and then the two-prong separately.
4. SEMILEPTONIC DECAYS OF CHARMED MESONS
The lepton spectrum associated with the multiprong sample is

shown in Fig. 10. The estimated background due to hadron misiden-
tification or heavy lepton production is also plotted. The back-
ground was scaled from measurements below threshold and the heavy
e+e- .... e:!:+~2prongs

3.0 3.99 < ECM < 5.20 GeV
MISIDENTIFIED HADRONS
HEAVY LEPTONS
~
.........
,.... 2.0
UI
2'
e
Q.
N
AI 1.0
Q5 1.0 1.5
bl cf
'tJ 'tJ Pe (GeV Ie)
Fig .. 10 The electron momentum spectrum for events with

three or more charged tracks for c.m. energies be-
tween 3.99 and 5.20 GeV. The curves are estimated
background from hadrons falsely identified as elec-
trons (solid curve) and from the production and de-
cay of a sequential heavy lepton of mass 1.91 GeV
(dashed curve).
lepton contribution was estimated for a lepton of mass 1.9 GeV and
a branching ratio of 30% into multiprong final states.
The electron spectrum contains information on the semileptonic

and the leptonic decay modes of the lowest mass 8 ,lO) charmed ha-
drons. Figure 10 demonstrates that semileptonic decays are much
more important than leptonic decays because the latter, being two-
body decays, would produce a peak in the electron spectrum around
1 GeV/c. This is in gross disagreement with the data which peak
around an electron momentum of 0.5 GeV with only few events above
0.7 GeV/c. To study the observed momentum spectrum in more detail
we consider the spectrum obtained for c.m. energies between 3.99
GeV and 4.08 GeV. In this energy region we expect DD-* and D*-*
D pro-
duction to dominate. The spectrum, corrected for the background
and the heavy-lepton contribution, is shown in Fig. 11.
The spectrum in Fig. 11 was fitted to three possible channels:

D 4- ev iT, D 4- ev K and D 4- ev K*(892). A V-A current was assumed
e e e
and the form of the spectra was taken from a paper by Ali and Yang 11 )
Note that the theoretical spectra are model dependent. These fits
gave a X2 value for 10 degrees of freedom of: 29.6 for D 4- eVeTI,
6.3 for D 4- eV- K and 2.e for D 4- ev- K* (892). The decay D 4- ev 'IT
e e e
can therefore be excluded as the sole semileptonic decay mode of
- K* (892) both give satisfactory fits
the D. D 4- ev K and D 4- eV
e e
to the data with D 4- ev e K*(892) giving the best fit.
The absolute cross-section for inclusive electron production

+ - +
e e 4- e- + X, where X contains at least two charged tracks and
any number of photons, is plotted in Fig. 12 as a function of ener-
gy. The data have been corrected for radiative effects. The back-
ground from hadron misidentification and the contribution from a
heavy sequential lepton have been subtracted.
376 B. H. WIIK
.
...---,
- >
u
ell
(!)
4.0
0) e+e- - e:!: + ~ 2 prongs
3.99 < ECM < 4.08 GeV
.0
C
'--'
o +ev K
3.0 o -evK (892)
III
0'1
C
0
L-
a. 2.0
N
1\1
+ 1.0
+,
ell
,+ 0.0
ell
+
ell
b
'0 l '0a... 0.0 0.5 1.0 1.5 2.0
Pe (GeV/c)
Fig. 11 The distribution of electron momenta, corrected

as described in the text for c.m. energies be-
tween 3.99 GeV and 4.08 GeV. The curves repre-
sent the expected spectrum of electrons from the
decay D + e~eK (dashed curve) and D + ev K*(892)
e
(solid curve).
The inclusive cross-section due to charmed particle production

can be written as:
+ -
Here a(e e + C.C.) denotes the effective cross-section for produ-
1 ]
cing the lightest charmed hadron stable against strong and electro-
magnetic decays. These particles might eigher be produced directly
or result from the cascade decay of excited charmed hadrons. The
cross-section a(e+e- + C.C.) was obtained by subtracting the cross-
1 ]
section for "old" hadron production (R = 2.5) and the cross-section
for heavy lepton production (m = 1.91 GeV) from the total cross-
section.
I I
2.5 r- (J (e+e-- e +~2prongs)

a)
2.0 f-
- 1.5 f-
f
-
-
r
.0
c:
b 1.0 f-
05 f-
g o. 'f
i
c.!) 0.25 f- 0'" (e+e--- e +~ 2 prongs)
z
b) 2 (J (charm)
I
~
u
~ 0.20 f-
a::
III
+++t+ I
u
Z 0.15 f-
....
0
n.
UJ
:::! 0.10 f-
~
UJ
VI
UJ
c.!)
005
c(
a::
UJ
~O.O
3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2
ECM (GeV)
~ig. 12 a) The cross-section for the inclusive production of

electrons plus non-showering track plus additional
charged tracks as a function of c.m. energy.
b) The average semileptonic branching ratio as a func-

tion of energy. The error bars are statistical only.
378 B. H.WIIK
Near threshold, where neutral and charged D production domi-

nates we find:
BR(J) .... e + X) = O.~O ~ 0.03
This should be compared to the value
B R (C -+ e T )() = O. .f 4 o. 03
obtained by averaging over all energies between threshold and
5.2 GeV. These values were extracted using PLUTO 3) data for the
total cross-section and the error quoted is mainly systematic.
Evaluating the branching ratio using the SPEAR data 2) on the to-
tal cross-section as input lead to an average semileptonic branch-
ing ratio of 0.08 0.03.
The semileptonic branching ratio can also be determined from

the fraction of inclusive electron events containing a second elec-
tron. Using this method we find BR(C + e v X) = 0.16 0.06 in
agreement with the value above. Note that this value is indepen-
dent of the charm cross-section.
The semileptonic branching ratio for charmed particles is,

therefore, substantially larger than the value of 4% predicted 12 )
from the weak decay of strange particles. This indicates that the
mechanisms responsible for enhancing the non-leptonic channels in
strange particle decays are less effective for charmed particle
decays.
5. LEPTONIC DECAY HODES OF THE T
Here we report results on the leptonic decay modes of the T

1 3)
particle . These results were derived from a measurement of the
following reactions:
+ -
1) e e + e + non-showering track
+ -
2) e e + W + non-showering track
+ -
3) e e + ew + nothing.
The results listed below are not yet final. The data on
reaction (1) are discussed above. Data on reaction (2) were ob-
tained by selecting events with a muon in one of the spectrometer
arms and a non-showering track observed either in the inner de-
tector or 1n one of the spectrometer arms. A charged particle was
called a muon if it had a momentum greater than 0.7 GeV/c, gave no
signal in the threshold Cerenkov counter, suffered an energy loss
consistent with that of a minimum ionizing particle in the shower
counter and penetrated at least 40 cm of iron for momenta between
0.7 GeV/c and 0.9 GeV/c and 60 cm of iron for momenta above 0.9
GeV/c. Muons between 0.7 GeV/c and 0.9 GeV/c were required to hit
the range chamber within 15 cm of the projected track direction.
A total of 14 events satisfied these conditions.
The observed muon yields were then corrected for the follow-
ing contributions:
a) hadron misidentification. The probability Ph~ that a hadron

satisfied the muon criteria were determined using hadrons
from the decay of the J/~ resonance and from measurements 1n
a test beam. Thus, experimentally, Ph~ was found to increase
from 2% at 0.7 GeV up to about 4% at 1.4 GeV and 8% at 2.5 GeV,
+- +- +-
b) e e + e e ~ ~ with the muons but not the electrons detected
+ -
c) e e + ~~y with the photons not detected. All these sources
were found to contribute a total of two events to the inclu-
sive muon sample.
Data on reaction (3) were obtained by selecting events with a

muon of momentum greater than 0.7 GeV/c 1n the outer detector and
an electron traversing either the inner or outer detector. A total
of 11 e~ events with Pe ~ 0.2 GeV/c and p~ > 0.7 GeV/c were found.
The background from misidentification and two photon processes was
estimated as 0.7 events.
380 B.H.WIIK
Are the observed events consistent with ~-e universality?

From reactions (1) and (2) we derive
~
N{e)
. -Ae-
A~
where A and A are the detection efficiences for electrons and
e ~
muons, respectively. We find B /B = 0.8 0.3, consistent with
~l e
e~ universality.
The momentum spectrum of the leptons, obtained by combining

the data from reactions (1) and (2) is plotted in Fig. 13. Note
that this spectrum is incompatible with the lepton spectrum ob-
served in the multihadron sample. The solid line in Fig. 13 was
computed assuming the leptons to result from the decay of a heavy
lepton of mass 9.9 GeV, a massless neutrino and a (V-A) weak cur-
rent. The agreement is good.
~~ ( nb/GeV )
0.5
0.25
0.5 1.0 1.5

Pe (GeV/c)
of- - +
Fig. 13 The lepton spectrum measured in e e ~ e(~) + T-.
The electron and muon data are combined for momenta
above 0.7 GeV/c.
If we assume T to be a lepton of mass 1.9 GeV then we can de-

rive the value of B (= B ) from the number of observed ew events
e W
We find
or
Be =0.2:!: 0.03
1 3)
This is in agreement with preV10US measurements and with theore-
14)
tical predictions .
6. SEMIHADRONIC DECAYS OF THE T
Both the production cross-section and the leptonic decay modes

of the T are naturally explained l3 ) by assuming the T to be a new
heavy lepton. Both the branching ratio and the momentum spectrum
of the decay leptons are consistent with expectations based on con-
ventional weak interaction theory (Fig. l4A). However, little is
known about how the T couples to normal hadrons. In a conventional
14) . .. .
theory ,the 1nteract10n w1ll be of the form V-A and lead to f1nal
states with a small ratio of kaons to pions (Fig. l4B). In particu-
lar, the decay T ~ V TI proceeds only by the axial current and is
T
directly related to TI ~ VW as depicted in Figs. l4C and D. The

b ranc h 1ng . can b e expressed 1 4
rat10 ) . .1n terms of the 1epto-
un1quely
nic branching ratio as
r (1:" - -+ \l"t' Tt- )

r( 't" -+"-te\>e)
with f = 0.137 M .
TI P
Similarly, the decay T ~ V P proceeds only by the vector part
T
+ -
of the current and its width can be uniquely predicted from e e ~
~ p assuming CVC.
382 B. H. WIIK
vl
"t-
A
e - . ....-
V. .V Il
t-
B
a a
d s
-cos9c -Sin9c
c ___"t.-___ W:: ~-
~~
9 gf~cosec
Fig. 14 A+B) The graphs for leptonic and semihadronic

decays of a heavy sequential lepton.
C+D) The relationship between TI and

T- -+ \! TI-
T
A characteristic feature of the standard weak interaction is

that decays involving strange particles are suppressed relative to
decays involving non-strange final states by tg 2 8. Apart from
c
small corrections due to phase space, the theory predicts
--------------- ==
We shall now discuss data obtained on the semihadronic decay

modes of the T. The results are very preliminary and they are
based on only a part of the available data.
We have determined the ratio of strange to non-strange par-

ticles in semihadronic T decays from a measurement of
where ny stands for any number of photons.
It has been shown above that the two-prong cross-section with

+ -
one electron predominantly results from e e ~ TT ~ (v ev)(v hny)
T T
with a small contamination from charm decays. The hadrons were
identified and measured using one of the spectrometer arems. The
electron was identified in either the inner or outer detector. We
find
O"(e+ K+ 'ttl( ) = O.Oi t 0.06

CT'(e+ 1t + 'n I )
i.e., on the average only 7% of all T decays yield a strange par-

ticle ~n accordance with theory.
This should be compared to multiprong events. We find for this

class of events a ratio
a-( e + K + ~ of p~ + Il\ l ) = O.24tO.OS

a-ee + 1'( + ~ i pl#rl9 + '"' r)
Since the charged multiplicity is of the order of 4, this yields
(0.90 0.18) Kjevent.
We next investigate the pion decay mode of the by searching

for final states of the type en + nothing. This final state re-
sults mainly from the process
384 B. H. WIIK
with a small contamination from other T decays with unobserved

particles or charm production.
The pion was identified and measured with the spectrometer

arms, the electron was identified both in the inner detector and
the spectrometer. To enhance the signal relative to the background
only pions with momenta above 1.0 GeV/c were considered. About
60% of all T + V n decays satisfy this condition.
T
The data were analyzed twice using independent programs. Both

methods found 4en events with p > 1.0 GeV/c.
n
14)
Theory predicts
1
.f. 04 'me = 4
1M!
"t
I
Using this as an input we expect to observe N(en) events with N(en)
= 2.55 B2.
e
With B
e
= 0.2 this yields 10.2 events. This is a firm
prediction of the theory; however, note that it depends strongly
on B .
e
The decay T- + V p- + V ~-TIo, where both photons from the
T T
decay are not measured, populate the en class. With B = 0.24 we
p
expect 1.75 en events from this source. There ~s a further 2%
chance to mislabel a hadron in the inner detector as an electron.
From the number of observed two-prong events we estimate 0.25 events
from hadron misidentification. Subtracting the two background events
(neglecting the charm contribution) leads to B B (0.004 0.005)
e n
compared to B B
e n
= 0.02 expected.
A prediction of the expected number of en events can also be

obtained from the number of observed ew events:
Alt
= .f-. __ '\J
2. A f"l
Thus from the observed number of 11 ew events we predict 11 en

events to be compared to the 10.2 events computed. Note that this
prediction is independent of B .
e
A further search for the pion decay mode was made by selecting
+
events of the type n- + charged track. Such events will result
from
Searching for this final state leads to an increase in statistics.

However, there is also a larger background in this class of events.
Besides T decays, also multihadron events with unobserved par-

ticles can contribute to such a final state. To reduce the back-
ground, only pions with momenta above 1.1 GeV/c are considered.
In the energy region between 3.99 and 5.2 GeV we e~pect 24

events from T -+ \\n and three events from T -+ \\P i.e., a total
of 27 events neglecting the leak-through from charm and multihadron
events. A total of 13 events was observed -- i.e., a 2.70 effect.
In the energy range from 4.52 GeV to 5.2 GeV we expected 17 events
compared to four events observed -- a 3.10 effect. However, note
that also these predictions depend on B2.
e
We have also searched for events of the type
eTe-~ "tF -+ (VT K)[(vTe'Y)+(V,:fl \ + (YT R'i')J-+

... K + dlQ1t~ed. t'to..cJ< + 4'lOtlUK9
Only one event with PK > 1.0 GeV/c was found. This corresponds to
a 90% confidence upper limit of B < 0.016.
K
We have also searched for T -+ V P by selecting final states
T

with the following topology: n + charged track + two photons. The
two photons were fit to a nO and events with a X2 < 6 were retained,
provided that both computed photon energies were above 50 MeV.
386 B.H.WIIK
DASP
PRELIMINARV
e+ e - -- + 0
Tt- Tt + lch
~ : ch = electron
10
all
5
a) 0 OL---'---"""10.f""'5~~~~""'1.0~---L----'1.5
o ~ ___ ~~~~~~~~L-_~
b) 0 0.5 1.0 1.5
+
Fig. 15 a) The distribution of M(n-no) observed for events
with the topology e+e- ~ nno + T+.
+
b) The M(n-no) distribution for events of the same
topology as above but with p n-n
+
> 1 GeV/c.
DASP
PRELIMINARY
0.5 < M ninO < 1 GeV
10
~ ch = electron
--~---
/
I
I
....
o /
o 1.0 2.0
+
Fig. 16 The momentum distribution of the (n-no) system
observed in e+e- + nno T~. The mass of the
nno system was between 0.5 GeV and 1.0 GeV.
388 B. H.WIIK
+
These events are plotted versus the mass of the n-n o system
in Fig. l5a. Events with an identified electron are hatched.
Events with the p band (0.5 GeV < M + 0 < 1.0 GeV) are plotted
n-n
versus the momentum of the nn o system ln Fig. 16. The momentum
distribution expected from the decay T ~ V P is shown as the dot-
T
ted line. Note the flat distribution above 0.9 GeV/c ln agreement
with the expectation. The enhancement at low momentum is pre-
sumably caused by multihadron events. To reduce the background,
+ 0
only events with a (n-n ) momentum above 1.0 GeV/c are considered.
The nn o mass distribution for these events is plotted in Fig. 15.
From these events we determine a branching ratio B
e
B
p
= 0.048
0.018.
With B
e
=
0.2 we obtain B
p
0.24 0.09 ln good agreement
with theoretical predictions.
The information on the T particle obtained from the DASP ex-

periment is summarized in the Table below.
Preliminary results on the T

r I
I
I
B
jJ
IB e = 0.8 0.3
e = jJ =
B B 0.20 0.03 I
T ~ vK-+
~ '" 0.07 0.06
T ~ vn
B
e
. B = 0.004 0.005
(T ~ vn not es tab lished)
B
p
= 0.24 0.09
BK < 0.016
REFERENCES
1) A more complete description of the various elements of the

detector can be found in publications of the DASP Collabora-
tion.
2) A.M. Boyarksi et al., Phys. Rev. Letters 34 (1975) 764.
3) J. Burmester et a1., Phys. Letters 66B (1977) 395.
4) S.L. Glashow, J. Illiopoulos and L. Maiani, Phys. Rev. D2

(1970) 1285.
5) DASP Collaboration, R. Brandelik et al., Phys. Letters 67B

(1977) 363.
6) PLUTO Collaboration, J. Burmester et al., Phys. Letters 67B

(1977) 367.
7) V. Luth et al., Phys. Letters 70B (1977) 120.
8) DASP Collaboration, R. Brandelik et al., Phys. Letters 70B

(1977) 132.
9) DASP Collaboration, W. Braunschweig et al., Phys. Letters

63B (1976) 47l.
DASP Collaboration, R. Brandelik et al., Phys. Letters 70B
(1977) 125.
10) G. Goldhaber et al., Phys. Rev. Letters 37 (1976) 255.

I. Peruzzi et al., Phys. Rev. Letters 37 (1976) 569.
E.G. Cazzoli et al., Phys. Rev. Letters 34 (1975) 1125.
B. Knapp et al., Phys. Rev. Letters 37 (1976) 882.
11) A. Ali and T.C. Yang, Phys. Letters 65B (1977) 275 and
private communications.
I. Hinchliffe and C.H. Llewellyn-Smith, Nuclear Phys. Bl14
(1976) 45.
F. Bletzacher, H.T. Nieh and A. Soni, Phys. Rev. l6D (1977)
732.
K. Kajantie, Phys. Rev. Letters 65B (1976) 69.
V. Barger, T. Gottschalk and R.J.N. Phillips, Univ. of
Wisconsin Report.
M. Gronau, C.H. Llewellyn-Smith, T.F. Walsh, S. Wolfram and
T.C. Yang, Nuclear Phys. B123 (1977) 47.
12) M.K. Gaillard, B.W. Lee and J.L. Rosner, Rev. Mod. Phys. 47
(1975) 277.
390 B. H. WIIK
13) For recent reviews on the status of the T see:

M. Perl in Proc. of the XII Rencontre de Moriond, F1aine
(1977)
G. F1ugge, DESY Report 77/35 and
Invited papers given at the 1977 Int. Symp. on Lepton and
Photon Interactions at High Energies', 25-31 August (1977),
Hamburg.
14) Y.S. Tsai, Phys. Rev. D4 (1971) 2821.

H.B. Thacker and J.J. Sakurai, Phys. Letters 36B (1971) 103.
DIS C U S S ION
CHAIRMAN: Prof. B. Wiik

Scientific Secretaries: H. Bohr, J. Field,
T. Weiler, S. Wu.
DISCUSSION
ATWOOD:
You mentioned the existence of a dip in the inclu-
sive e+e-~ X cross-section distribution at Ecm 4.3
GeV. How deep is this dip and does it give any trouble
to the existence of heavy leptons?
WIIK:
I did not mention any dip in the inclusive electron
cross-section. We do not have sufficient data in that
region to decide that. The total cross-section e+~~
hadrons does indeed have a dip in this region. However
note that a heavy lepton with mass 1.91 GeV contributes
less than one unit of R at 4.3 GeV and hence it might
be difficult to get conclusive evidence for or against
a heavy lepton from such data.
ATWOOD:
The dip in R suggests a decrease in charm product-

ion at Ecm = 4.3 GeV. Wouldn't this be the best place
to get a "purer" sample of heavy leptons? One could
then look for changes in the inclusive electron spectrum
and mUltiplicity both inside and outside the dip region
to convince the "doubting Thomas's" among us of the
existence of the heavy lepton?
391
392 DISCUSSION
WIIK:
Yes, this is a good place to look. For the "doubt-

ing Thomas's", I point to the fact that the low multi-
plicity events show no associated K's, while the high
multiplicity events have associated K's, i.e. the weak
current producing the high multiplicity class likes to
couple to strangeness. The weak current producing the
low mUltiplicity class does not.
SCHOPPER:
The Heidelberg collaboration wants to run in this

region for specifically the points Atwood has raised,
while the DASP people want to run at Ecm = 4.4 GeV to
increase the statistics on the F meson.
SANDA:
Professor Rubbia told us yesterday that we should be

careful about three standard deviation effects. But let
us believe, for the time being, that "C goes to 'tJ-c It
is not seen. If the naive prediction for ~ goes to ~~
is wrong, the '"t' goes tO~f prediction, and for that matt-
er all semileptonic decay predictions, cannot be trusted.
We cannot, then, conclude that the semileptonic decay of
~ has low charge multiplicity. In this case your crit-
erion for saying that low charge multiplicity events in
e+e- .... eX come from the ~ becomes suspect. It seems
to me that experimentally one cannot extract ~ events
from e+e--+eX without first making an assumption about
the multiplicity of ~ decays. Thus, before one concludes
that there is a new type of weak interaction in ~ decay,
one must figure out a way to extract the "'C signal from
e +e - -+ eX . .
~n a model ~ndependent way.
WIIK:
The problem is not at all serious. If we assume

reasonable properties for charm decay, we find that
charm cannot explain the two-charged prong events. So
the two-charged prong events in excess of the meager
charm background must be attributed to some other source.
Also note that our observed value for 1: -+ yTt is only
2.7 a away from the expected value.
DISCUSSION 393
WHITE:
When you discuss the data for e+e--op e++ one-charged

track, you said that you had just doubled the statistics
and now had 80 e 1t events and 2 e \<.. events. Does your
plot showing the energy dependence of~-type events
include these new data? The reason I ask is that we
have seen that R goes through large variations over this
energy range and drops to about 2.5 at 4.3 GeV. I
wonder when the statistics on"'t -type events will be good
enough to show that they do not track with these varia-
tions in R.
WIIK:
We will not have much better statistics in the near

future. To answer this question you have to go to a
large solid angle detector like DELCO at SPEAR.
NEW PARTICLES OR "WHY I BELIEVE IN QUARKS"
A. Martin
Theory Division, European Organization for Nuclear Research
Geneva, Switzerland
1. INTRODUCTION
One of the most striking events in the last three years has been
the discovery of the new particles. By new particles I mean the J/~,
~/, X's and also the charmed bosons. There are other new particles
such as the heavy lepton T or the "baryonium states" or the candidate
for charmed baryon that I shall leave aside. The reasons I became
interested in this particular group of particles are:
i) that it is difficult not to be excited by such discoveries,

even if your field of interest is miles away, namely rigorous
bounds on scattering amplitudes and, at the other extreme,
rigorous inequalities on non-relativistic systems;
ii) that it appears that the members of the J/~ family can be des-
cribed as a non-relativistic system of a charmed quark-
antiquark pair with great success (of course, the non-
relativistic Schrodinger equation can result from the
reduction of a relativistic equation), with, of course, a few
exceptions; I realized that this is a sufficiently simple
situation to allow investigation by rigorous methods, within
395
396 A. MARTIN
Table 1
Elementary particles <November 1977)

Table taken from lectures I gave at the University of Paris
in 1957. Notice that the N*3 already discovered 3 was not
considered as a particle.
Name Mode of Disintegration Lifetime

(seconds)
Photon Stable
~ j""C<O"
Neutrino Stable
Stable
-6
S Meson].l ].I "'e +v+v 2,2 x 10
(ou V + v?)
+ +
11 "'].1 + V 2,5 X 10- 8
Mesons 11 11- ... ].1- + V
11 ... 2y T < 10- 15 (7)
11 + 11 + 11 ;
.' {
11 + 11; -8
].I + V; 1,3 0,3 x 10
+
].1- + y + v; ].I + rro + v;
Mesons K e + y + v; e + rro + v.
KO = KO + KO ... rr+ + rr- 0,85 X 10- 10
I 12
P stable
Nucleons N ... P + e- + V (ou v?) 1110 220
B
A
R
Y
o 1:;+ ... {' P + rro 0,8 X 10- 10
N N + 11-
Hyperons
S 1:;- ... N + rr- 0,8 x 10- 10
1:;"'Ao+y very brief
-10
1 x 10
NEW PARTICLES 397
the framework of the Schrodinger equation, the sensitivity of

the spectrum and decay mechanisms to the details of the inter-
action, and try to get, through a few theorems, some insight
on how much model-dependent are the results; on the other
hand, I shall not touch at all fundamental questions such as
the confinement of quarks; I think that for these models it
makes absolutely no difference whether quarks are exactly
confined or approximately confined, escaping after 10 20 years;
iii) that Professor A. zichichi gave me the order to lecture about

new particles instead of working on "uninteresting" problems
of mathematical physics;
iv) that I have taught before on the subject of new particles, as

shown in Table 1, which comes from lectures given in Paris in
1958.
My plan will be the following. First I shall try to describe

in a minimal way the experimental situation concerning the members
of the J/~, ~' family and, in particular, discuss the attribution
of quantum numbers. You will see that it is not yet possible to
make all attributions without theoretical prejudice. In particular
this is what happens for the X's (3550, 3510, 3450, 3410, and
2850 GeV). I shall discuss what can be done experimentally to im-
prove the situation, a thing which seems to me very essential if
one wants to consolidate a theoretical model which, after all, has
not a very strong basis.
I shall briefly discuss "open charm" and also possible higher

states well above the charm production threshold.
Then I shall try to review some of the existing models and

present to you a few theorems which help to understand the level
scheme and the widths of the new particles.
398 A. MARTIN
Finally, let me say that I find it very difficult to present

all references on the subject. I admire, for instance, B. wiik
1)
and G. Wolf for the very thorough job they made in lectures
at Les Houches last year, but please do not expect that of me.
Other useful general references are V. Luth's lectures 2 ), and
J.D. Jackson's talk at the Budapest Conference 3 ), and also the
experimental talks of Litke, Goldhaber, and Timm at the same
Conference'+)
2. THE J/~, ~' FAMILY AND THEIR DAUGHTERS
2.1 IN and ~'
The J/~ and ~' are very narrow resonances produced in e+e- col-
lisions at 3095 and 3685 MeV, respectively. Their total width is
estimated by the following method: one assumes that the cross-
section has a Breit-Wigner form. Integrating over energies after
subtraction of the background one finds the production width which
is r e , independently of the energy resolution of the machine:
Similarly, the ratio r e Ir tot can be obtained by
J~~+p--(t:')dE'
Jo-T (')dE'
where a~+~_ represent the cross-section for producing a muon pair

(assuming that the e+e- width is equal to the ~+~- width).
After elimination of background and radiative effects, one

finds
NEW PARTICLES 399
re = r r tot
II
J/lJJ 4.8 0.6 keV 69 1.5 keV
t/J' 2.1 0.3 keV 228 56 keV
In composite models the leptonic width is proportional to the

square of the quark-antiquark wave function at the origin. We shall
come back to that question in Section 4: the comparison of the two
widths gives some indications on the potential.
We now want to know the quantum numbers of the J!t/J and t/J'. In
both cases it is possible to show that the resonant amplitude in
e+e- + ll+ll- interferes with the Bhabha term, destructively below
the resonance energy, constructively above. This shows that J!t/J
PC
and t/J' have the same quantum numbers as the photon J = 1
Next, we come to isospin, G parity and SU(3) classification.

When one looks superficially at the pionic decay modes of the J!t/J,
one sees that it produces both odd and even numbers of pions. How-
ever, odd numbers are dominant. Even numbers can be accounted for
by a second order electromagnetic effect, i.e. production of a
virtual photon by the resonance followed by hadron production.
This mechanism is confirmed by measuring the ratio of a(n pions)
to all +ll - on and off the resonance and showing that for n-even this
ratio does not change, while it changes a lot for n-odd. The con-
clusion is that G parity is conserved in purely hadronic decays of
the Jl'4J and is odd. From the relation G = C(-l) I we deduce that
isospin is even. But the decay J!t/J + pp, with a probability of
0.22 0.02% is too large to be explained by QED effects, and is
only compatible with I =0 or I = I, and hence I = O. Another
proof is given by looking at the TIP decays of J!t/J. One finds that
400 A. MARTIN
a(n+p- + n-p+) ~ 2a(nopo), in agreement with I = O. Finally, there

is the question of the SU(3) classification of J/~. One uses the
fact that it is forbidden for an SU(3) singlet with odd-charge
conjugation to decay into two members of the same SU(3) multiplet
[and also into two objects belonging to multiplets whose neutral
. . 5)]
components have the same charge conJugat~on One observes that
the decays to K+K-, KOKo are at least 20 times less abundant than
the decays to K+K-*, KOKo*, KO*Ko**. Another confirmation of the
singlet assignment comes from comparing pp and AX, which should
have equal rates except for phase space. The same holds for n-p~
and K-K+*. Data indicate that the maximum amount of octet compon-
ent is in amplitude 12%.
In the case of the ~' the situation is quite analogous. First

of all, % of the decays 80 to IN n+n- and 1/6 to IN nOno. This
indicates that: i) G parity is odd; ii) I = 0 from the comparison
of n+n- and nOno rates. This is also confirmed by the observation
of J/~ + n decay, and of pp decay. Concerning the SU(3) assignment,
all one can say is the following: if it is a singlet, the decay to
K+K- is forbidden and is observed to be less than 5 x 10- 5 However,
we have no data on the K*K modes for comparison. But for pp (al-
lowed by singlet) we observe 2.3 0.7 x 10- 4 So it seems reason-
able to take ~' to be a singlet part. Then, of course, one has to
explain the ~' + J/~ + n decay by the singlet admixture of the n.
2.2 The X states
It is difficult at this stage to keep completely away from

theoretical considerations. In spite of my desire to remain neu-
tral for the time being with respect to the theoretical interpre-
tation of the new particles, I have to mention that before the X
states were discovered, two groups6,7) had theoretically predicted
the existence of intermediate states between the ~' and the J/~,
NEW PARTICLES 401
in a model in which these particles are bound states of charm-

anticharm quarks. In fact, the prediction of charm-anticharm S
states was made by L. Maiani at this school in 1970 8 ) and later by
Appelquist and Politzer 9 ). The reason why a charm-anticharm pair
has been chosen rather than any other particle-antiparticle system
is the narrowness of the J/~ and ~' states. By the Okubo-Zweig-
Izuka rule, the J/~, ~', if made of quark-antiquark, have their
decays into ordinary hadrons strongly suppressed, because quark
diagrams are unavoidably disconnected. Do not ask me the justifi-
cation of this rule.
c
3iT's
The predictions, made mostly with a potential V

+ A + Br, are as follows.
One has singlet and triplet states depending on how one com-
bines the spin of the quarks. The charge conjugation is given by
c= (_,,)1-+5
parity is given by
L..
p =-(-1)
We get then this picture with the indication of J PC :
1..'=2.,
J... =0 l =1 __..-""'--___--,
.....-"
r-----~:r.
-------"-
.. - - - -
_~(
___
0
~
~") '3F-
......""""-::--_ _
2+1'
,-+
"!'1--c=-:----- ----~
1+1 /+-
----OL.+7:.,.::-
402 A. MARTIN
We shall discuss later the relative position of the levels,

but let me say that with the type of potential that was used, the
L = 1 levels were predicted as lying between the ~' and J/~.
Intermediate states have indeed been observed in several ways:
i) monochromatic peaks in the inclusive photon decay spectrum of

the ~';
ii) two photons followed by leptonic decay of ~;
iii) one photon plus hadrons.
In (i) and (iii), decays of the ~' to three states + one pho-
ton have been observed, the energies of the states being 3.41 GeV,
3.50 GeV, and 3.55 GeV. In (ii) an additional state was found. In
fact, there was an ambiguity to resolve: in the decay ~' + y + X,
X + Y+ J/~, which y is emitted first? This has been answered be-
cause as X is in motion, the second y is not monoenergetic. The
states found are 3.41, 3.45, 3.50, and 3.55 GeV. The 3.45 state,
considered as doubtful for some time, partly because of low statis-
tics, partly because of its absence of decay into hadrons, has re-
cently been confirmed by PLUTO.
Here the attribution of J PC quantum numbers is much more dif-

ficult, that is if one does not want to appeal to theoretical pre-
judice. One can, of course, safely say that C = +1, since ~' +
+ X + y,.and X + Y+ ~ for all four states.
Information on spins and parity can be obtained by looking at

single and double y-ray distributions and also hadron decay and
correlations between the first y and hadronic decay. Another easy
attribution for the states 3410, 3510, 3550 is the attribution of
G parity: it is seen that only even numbers of pions appear in
pionic decay and so
NEW PARTICLES 403
From C = +1 and from the formula

r
G- =C (-1)
we 'see that the isospin of these three states (if it is meaningful!)

has to be even. However, in the decay ~' + y + X we have I~II ~ 1
to lowest order in electromagnetic interactions and therefore,
since I = 0 for ~', we attribute I o to 3410, 3510, 3550. How-
ever, we must be careful, because we have no guarantee that isospin
is conserved in the decay. Fortunately, charge conjugation is good
enough. If a state with C = +1 decays into n+n- or K+K- we conclude
that:
i) it has natural spin parity 0+ 1- 2+ etc.;

++
ii) however, C = +1 eliminates odd angular momenta so only 0
2++ 4++ etc. are allowed.
The I 3410 and 3550 states I are seen to decay into n+n- and
K+K- and therefore they have I Jeven, P = C = +1 I.
The state 3510 does not decay into n+n- or K+K-. This is not
..
a very severe restr1ct10n. . excludes 0++
As f ar as I can see, 1t
++ ++
2 4 etc., but no more.
The second restriction comes from the angular distribution of

the y's. The simplest thing to do which has been done experiment-
ally is to look at the angular distribution of the first y from
~' + y + X. Only ~ prediction is model independent: if ax has
spin 0, the angular distribution of the y with respect to the beam
is
1 + cos 2 e
However, finding that the distribution is so does not prove

that it is spin o. What is found in a fit 1 + a cos 2 e is
404 A. MARTIN
3414 ex. = 1.4 0.4

3508 ex. = 0.25 0.5 or -1 0.5
3552 ex. 0.22 0.4
3454 ? ? .
The conclusion is
3508 and 3552 have J # 0
Hence
JPC(3552) 2++ 4++ etc.
This is as far as we can go being completely rigorpus. Let

us notice, however, that on the basis of general positivity require-
ments we have ex. ~ 1. If one finds that ex. is exactly equal to unity
it will be only by a dynamical accident (vanishing of the amplitude
for producing a helicity one X) that a spin J # 0 could simulate a
spin 0 and therefore it is likely that 3414 has spin 0++.
If you simplify the rules of the game by deciding that you

want to identify the four states with the four states 0++ 1++ 2++
0-+ predicted by the charmonium model, you find, following Chanowitz
and Gilman 1 0 )
2.++ 040- '3 S"S' 2-
r ...-- JeV'~:J Sri

-
O .... '34- cL,. 1'l =F O
1.+ .... '"3 S'o g
The left-over has to be
0
-+ ~ '34$0
For the first three states, this kind of assignment is consist-

ent with the widths predicted for an electric dipole transition
(2J + 1)k 3 x const. where k is the photon momentum, knowing that
NEW PARTICLES 405
the space wave functions of the three states, in a composite model,

are the same.
However, my personal opinion is that we should not be contented

with this situation and that every effort should be made to obtain
a model-independent determination of the spins of these four states.
Sooner or later we shall know the angular distribution of ~' ~
-+
~ 3450 + Y and this might exclude the 0 assignment in favour of
a proposa 1 0 f Harar1 .11) . .
to make th1s a D state, w1th J = 2 ,or
consolidate J 0, with, however, the possibility of accidental
vanishing of an amplitude which would mimic J = O.
12)
It has been suggested by various people to study correla-
tions, either in the chain
or
spin-O mesons
The second procedure, when it can be applied (i.e. for 3414 and
3552) is the most advantageous one because it minimizes the number
of amplitudes since the final particles have spin-O. This is what
has been pointed out by Kabir and Hey in particular.
They obtain an expression for the correlation function
where ey and eM are the angles of the y and the mesons with the
beam and ~M is the relativ~ azimuthal angle. These are complicated
expressions depending on J+l parameters which are the J+l amplitudes
for producing a X with helicity 0, 1, . , J. Irrespective of the
value of these parameters the spin is fixed by the (perfect!)
knowledge of W2.
406 A. MARTIN
In the case of yy correlations the situation is not as nice.

In that case Kabir and Hey show that "accidents" can occur produc-
ing a completely isotropic angular distribution of the second
photon for fixed direction of the first photon and simulating
therefore J = 0, even though the spin is 1 or 2. However, Kabir
and Hey, limiting themselves to J ~ 2 show that the difficult
cases can be resolved by looking at the lepton pair of the final
~ decay. If, for instance, the distribution of the second y is
isotropic, the correlation function between the first y and the
muon pair, averaged over azimuthal angles, takes the form
( 1 + Ai loS ~ & ) ( 1 .... Ar (.0,'2. &t )
if JX =: 0 At'- :::' 1
JX := 1 At" := - 3/r:
1X 2 Ap :: - i/1
Therefore, with sufficient statistics a determination of the

spins of the X's, free from theoretical prejudice, is possible.
Finally, let us indicate 13 ) that a spin larger than 2 would

unavoidably show a characteristic pattern in yy correlations (with-
out study of the lepton pair).
One could also ask oneself if the polarization of the e+e-

beam, which is parallel to the magnetic field and occurs only in
certain circumstances, could be used. By a theorem of Bjorken 14 )
the answer is negative. In the ultra-relativistic case the only
thing the virtual photon remembers is that its linear polarization
is perpendicular to the beam. So the unpolarized cross-section for
a beam along the z axis, for any exclusive process is
NEW PARTICLES 407
where a and a correspond to linear polarizations of the photon

x y
along the x and y axes. From this formula we get
where a corresponds precisely to electrons and positrons polarized

x
up and down along the magnetic field. So the polarization of the
beam introduces a new azimuthal dependence but gives no new in-
formation, at least for a 4n detector. It can help in practice
because two cones are excluded from the detecting apparatus.
Now we come to another X state which lies below the J/~ and
has been observed at DORIS 4 ). In the e+e- + 3y spectrum a peak
is observed at 2850 MeV in the yy system after subtraction of the
contributions of ny and n'y. Though this state has not been ob-
served at SLAC, evidence for its existence is accumulating at
DESY. Naturally, the classical arguments of Yang apply to this
state which decays like X +yy, and therefore it has J ~ 1, and
C = +1. It is tempting to attribute the quantum numbers 0-+ to
this state, but not yet fully justified.
Finally, a new state 4 ) has been observed very recently at

SPEAR, produced directly in e+e- collisions at 3772 GeV, with a
width of 28 5 MeV, and a leptonic width of 370 100 eV. This
state was not seen before because it is in the radiative tail of
the ~'. As we shall see, this is slightly above the charm-
anticharm production threshold and indeed this new state is seen
to decay in DD. This state could be interpreted at 1 (photon
quantum numbers), possibly a L =2 state in a composite model.
3. THE CHARM SECTOR 4 )
New states have been observed around 1870 GeV. In fact, they
have been looked for very actively before they have been discovered,
for two reasons.
408 A. MARTIN
i) The Glashow-Iliopoulos-Maiani mechanism for suppressing

strangeness changing neutral currents requires the existence of
charmed particles. In terms of quarks it requires the existence
of a fourth quark:
B S Q y C
u 1/3 0 2/3 1/3 0
d 1/3 0 -1/3 1/3 0
s 1/3 -1 -1/3 -2/3 0
c 1/3 0 2/3 -2/3 +1
ii) The second reason is that it is extremely tempting to regard

the ~ and X states as bound states of pairs of new quarks. The
small width of the ~' indicates that the threshold for the produc-
tion of charmed-anticharmed mesons should be close to or above
3700 MeV.
In the quark model the three charmed mesons are the
The structure of the weak charged current in the GIM is
where 8 is the Cabibbo angle.
This leads to the selection rule
for the favoured transitions, while transitions

NEW PARTICLES 409
are suppressed and
AQ: A C. a - A5
is strictly forbidden.
So one expects
while
t> +-> '1t - 1l" + 'It i- ,
is suppressed, and
is forbidden. It is worth noticing that in the "old physics" a

state like K-n+n+, which has at least isospin 3/2 is "exotic",
i.e. could not be made of a quark-antiquark pair, while K+n+n- is
not exotic. So the n+ decays will have a very clear signature, and
could not be mistaken with the decay of a new excited K. On the
contrary, the decay of the F+ (cs) has not an easy signature. It
could decay into a non-strange system like n + nn or KK + nn's.
What has been observed is the following:
i) States near 1.86 GeV:

+ - + + + -
neutrals decaying into K-n +, K-n-n n with mass at 1864 2 MeV,
width < 40 MeV. I do not want to describe the tricks of the
analysis in which one difficulty is the imperfect separation
between K's and n's; the initial e+e- production energy was
4.03 GeV but new data have been obtained around 3.77 GeV.
charged particles decaying into K+n-n-, K-n+n+ but not K+n+n-

or K-n+n-, with mass 1869 2 MeV, width < 40 ~1eV; it is
natural to identify these states with nOno, n+n-. Notice the
purely "exotic" decay of the charged n's which clearly ex-
cludes excited K*'s.
410 A. MARTIN
ii) At 4.03 GeV e+e- energy one observes peaks in the recoil spec-
trum of the DO and D from which one can infer the existence of
DO*, mass 2003 MeV, width < 40 MeV, with decay DO* + TIoDo and yDO
and D*+ at 2010 MeV, with D+* + TI+Do. There are other peaks but
it has been shown that these can be accounted for by kinematical
reflections and are caused by the very small Q value of D* + TID.
Decays into n*D* have also been observed.
Now one has to attribute quantum numbers. The small mass dif-
ference DOD+ makes us put it in an isospin doublet. The spin is
tentatively taken to be zero because a spin-l would, in general,
show decay correlations which are not seen.
The decay D*o + y + DO shows that not both DO and D*o have
spin-O. Spin-O is favoured for DO by the angular distribution in
e+e- + 3.772 + DOnO 15), hence we are tempted to attribute spin-l
to D*o. If, according to the theoretical prejudice, DO is 0-, D*o
is 1- because of the existence of the strong decay D* + D + TI.
3.1 The question of parity violation
The decay D+ KO + TI+ very recently observed, indicates that

short
if parity is conserved D+ belongs to the natural spin-parity se-
+ - + + - + + .
quence 0 1 2 etc. However, the decay D + K TI TI 1S also ob-
served and one can repeat exactly the same exercise that was carried
out 22 years ago, comparing the decays e+ TI+TI-, T + TI+TI-TIo.
+ - + +
Notice first that in D + K TI TI the subsystem TI+TI+ is only
allowed to have natural spin-parity 0+ 1- 2+. If D+ has spin-O
(favoured by production data) the relative angular momentum of
K-( TI +TI +) 1S
. equal to the sp1n
. (+
TI TI + ) and hence D+ should have
parity -. For higher spins one can compare with the distributions
calculated by Zemach which give the minimal structure for a given
assignment. The conclusion is that the distribution of momenta of
the K's is uniform and 1- 2+ etc. are excluded. We are left with
NEW PARTICLES 411
spin-O, for which there is an inconsistency, i.e. OT from two-body

decay, 0- from three-body decay. Parity is therefore violated.
Pushing the analogy with the 6-T puzzle, we can ask ourselves
if there is mixing of the DOno as in the KOKo system. More exactly,
is the mixing so fast that produced DO,s decay in K+n- as well as in
K-n+? The test consists in looking at the decay products of e+e- ~
~ DOno. In the absence of mixing we should observe only K+K-n+n-.
For 100% mixing we should observe e+e- ~ DOno, where D is the short-
s s s
lived combination (notice that e+e- ~ one virtual photon ~ DODo is
s L
forbidden by CP conservation), and events with final states KKn+n+
should be as frequent as K+K-n+n-. Experiment by the SLAC-LBL group
indicates that the mixing is less than 16%.
3.2 The unfavoured decays
We have now upper limits on DO ~ n+n-/Do ~ K-n+ and DO ~
~ K+K-/DO ~ n+K- of 7% compatible with the value of the Cabibbo

angle")
Evidence for the existence of the F meson has been obtained

at DORIS. One assumes that F is produced as e+e- ~ F-*F+, F+ ~
~ nn+. One tries to look for the production of n's (from the F
decay) accompanied by a soft y (energy < 140 MeV). At 4.4 GeV
production energy a clear signal is detected. For more details,
see the lecture by B. Wiik 16 ). The masses are 2.03 0.06 for the
F and 2.14 for the F*.
4. A THEORETICAL DISCUSSION OF THE J/~ PARTICLES
It is out of the question to make a complete presentation on

the whole literature on the subject. Following the example of many
others, I shall unavoidably concentrate on the a~pects to which I
have contributed.
412 A. MARTIN
The interpretations in which these var~ous states are colour

states have been essentially universally abandoned. Everybody
seems to agree that the J/~ particles are bound states of a charmed
quark and a charmed antiquark. Where people differ is on what
these quarks are: are they physical particles interacting via a
potential which lets them escape so slowly that they are very diffi-
cult to observe? Are they permanently confined by infrared gluons
which can possibly be simulated by a bag with surface and volume
energy, or by strings? Are they "mathematical" objects? I shall
not at all try to answer these question about which people fight
with great verbal energy and which sometimes are pure questions of
semantics: I do not see a great difference between mathematical
quarks and permanently confined quarks. There is only one fact on
which I think everybody will agree: it is that we need quarks to
explain many phenomena in high-energy physics. The most spectacu-
lar case is that of the new particles, but there are many other
things such as the absence of exotic states, the values of the hy-
peron nucleon cross-sections (1% check by Steinberger of the Lipkin
sum rules), the ratio close to 3/2 for NN/nN, total cross-sections,
and so forth.
Once you agree that the "new particles" (1975-76) are made of
a quark-antiquark pair, you have to decide what is responsible for
the binding. There you can start at various levels, from a funda-
mental point of view to a purely phenomenological point of view.
The fashionable point of view consists in saying that quarks are
bound by coloured gluons, but all you can do is to compute the
single gluon exchange which produces a kind of relativistic Coulomb
potential in which you can even decide that the coupling constant
dies logarithmically for r + 0 because of asymptotic freedom;
here at least one has a relatively clear prescription for the
choice of the vector character of the potential, and this is rele-
vant when one goes to the non-relativistic reduction to know what
will be the spin effects accompanying the "Coulomb" force. However,
NEW PARTICLES 413
nobody knows how to compute the "confining" force, its spin depend-
ent effects, etc. So, depending on what you assume in your rela-
tivistic equation as interaction, you get different results in its
What I want to argue is that it is as well to start non-

relativistically from the very beginning, or if you want, to justify
laziness, precisely because of our ignorance and, also, because it
happens that the spectrum is fitted by a relatively high quark mass,
about 1.6 GeV, and that quarks are indeed not moving too fast.
The standard model proposed first by Applequist, De Rujula,

Politzer and Glashow 6 ) and Eichten, Lane, Kinoshita, Gottfried and
Yan 7 ) has a central potential of the form V = -(4/3)(a s /r) + A + Br.
This potential produces a sequence of levels (using the hydrogen
notation):
,C-
3$
< -+
~
o 3P
3-~
-i --
3D
( 2-- l.- r
25 ( ,0
- ....
r-
2++
~
1 +~
I
i +-
ott
,,--
"'5
~ 0
-+
which splits into singlet and triplet states. We have indicated

J PC of t hvar~ous
e' .
states us~ng
= _ (_ i) L
.: (_-i)I..+S
414 A. MARTIN
The spectrum appears to be in qualitative, and almost quali-

tative, agreement with the data. The most spectacular prediction
is that of three P states. I was the witness of a talk given by
K. Gottfried in Dijon before the discovery of the C = +1 states.
He said that the model would be definitely ruined if it was con-
firmed that these states were absent. Another spectacular predic-
, ,
t10n by var10US groups 17,18),
1S t h at 0 f t h e tr1p
'1 et 3D, 1 state
at 3.77 GeV which has just been recently discovered, which we
shall call ~". One could argue that the agreement is too good.
It is true that there is an element of luck. However, what is
significant is that within 10 MeV of the ~' a new state is observed,
which can be directly produced by one photon in e+e- collisions
(see below a discussion of the production mechanism).
The prediction of the pseudoscalars is also a success, assum-

ing that 3454 and 2830 are really that (this attribution may be
destroyed by a study of the angular distribution) but the singlet-
triplet mass splitting is much larger (~ 300 MeV) than what can be
predicted by taking a simple-minded relativistic interaction and
carrying the non-relativistic reduction 18 - 20 ). However, let us re-
member that the true interaction is not known. Another problem
with the "pseudoscalars" is the magnitude of the transition V -+
-+ y + P. If the central potential is dominant or if the spin-spin
interaction is concentrated at the origin (as it is in atomic phys-
ics) the magnetic dipole transition is favoured because the space-
wave functions of the two states are almost identical. So what is
needed is a large, extended spin-spin force. Forces of that type
h ave b een cons1'd ered by Schn1tzer
,19)
.
Following a question by H. Lipkin, let me make a personal re-

-+ -+
mark on the question of the splitting by terms of the type 01.02.
-+ -+ -+
Usually, a term (01.02) 6 3 (r) appears in the Schrodinger equation.
This term is only well-defined in perturbation theory. An approxi-
mation to 6 3 (r) will be
NEW PARTICLES 415
the delta function corresponds to the limit R + O. However, per-

turbation theory is only applicable for R- VO x R2
3 1, i.e. VoR- 1
much less than 1. So R cannot be made arbitrarily small. In atomic
physics there is no problem because we have a natural cut-off due
to the extension of the nucleons. Here not. Taking R very small
is only allowed if one solves exactly the Schrodinger equation.
Then one finds i) that in the triplet states the interaction pro-
duces a negligible energy shift; ii) that in the singlet state
large shifts can be produced and the wave function is radically
changed so that the favoured magnetic transition between 3 2S and
3 lS is reduced, while the orthogonality between the 12S and 3 lS
is destroyed, making a magnetic dipole transition possible.
With these restrictions the model is a success. There are, in

fact, other successes such as the transition rates ~' + X to the
supposedly P triplet states. Another success to be discussed later
is the description of the state at 4.028 GeV (for which, however,
alternative interpretations have been proposed, such as molecular
charmonium).
Let us notice also that the non-observation of the singlet P

state is relatively well understood. It cannot be produced directly
in e+e- collisions because it has the wrong parity. It cannot be
producted by photon cascade from a state directly produced in e+e-
collisions such as ~' or ~I/. A detailed analysis of its possible
21)
production and decay has been made by Renard
Before going to the higher states, we want to discuss the sensi-

tivity of the model to details. In the purely central approximation
the level sequence appears to be very stable.
416 A. MARTIN
~f
What Harald Grosse and I have proved 22,23).~s .
th~s: ~
Theorem I
If
and
(#I. )
and as well
i.e. the X states with C = +1 lie between the S states. Remember

that for a pure Coulomb potential E2P = E2S ' For a pure harmonic
oscillator E2P = ~(E2S + E1S ) (here I keep the hydrogen notation).
Theorem II
If, in addition to the previous conditions, we have
-
d
Il.r r
we get
where E3D represents the lowest L = 2 level. Notice that in the

case of a pure harmonic oscillator potential E2S = E3D , and pre-
cisely the left-hand side of (**) is zero.
NEW PARTICLES 417
For those who think that these two theorems are not very trans-
parent let me indicate that they tell you that in
v::-- + 'B + Cr
you can replace -(4a/3r) by any superposition
-J
_I
o
f(oC)rllidol f(O<) >0
In particular you can weaken the "Coulomb force" if you like to have
asymptotic freedom, and take, for instance, instead of -(a /r)
s
Then the level sequence
remains unchanged. Potentials rising faster than r2 may exchange

the order of the 2S and 3D levels.
These conditions are, of course, only sufficient conditions.

My interpretation is that the level sequence shows great stability
as soon as one admits the idea of a confining force. Naturally,
one can build counter examples by constructing ad hoc potentials,
but these counter examples do not seem very natural physically.
Let me say a few words on the methods of proofs of these

theorems.
What plays a major role is: i) the nodal structure of the

radial wave functions -- the IS and 3D states have no nodes, the
2S state has one node; ii) the fact that there are extreme cases,
Coulomb and harmonic potentials, for which the solutions are known;
iii) the use of continuity of the energy levels with respect to a
parameter in the potential; iv) the virial theorem which says that
the kinetic energy is given by
418 A. MARTIN
" Sr
2..
~ u.1 elr
ei,.r
while the Eotential energy is given by
jV",ld,r
Details on the proofs may be given in the discussion section.
There are, however. other quantities which depend on the wave

functions and therefore, indirectly, the potential. This is the
case for the leptonic width, of the 1 states which, for coloured
quarks reduces to
where M is the mass of the bound state.
The decay width of the ~' is 2.1 keY, while that of the J/~ is
4.8 keY. What potential will produce such a ratio? In a harmonic
oscillator potential the wave function at the origin of the 2S state
is larger than that of the IS state. This is a clear indication
that a pure harmonic oscillator is not acceptable. In a purely
linear potential the magnitude of the wave function at the origin
is the same for all = 0 levels 2") ,for it is given by
00
4'Tr'ly,(o)l2.: [u.'(o)I2.:: J t1V u't(r)tJ.,r

D c:l-r
and since all wave functions have the same normalization and dV/dr
const. all the ~(O) are equal.
Recently I have proved a more precise theorem 25 ):
Theorem III
1) If the full potential is a concave function of r

NEW PARTICLES 419
2) If the potential is convex
In particular the favoured combination
-:3 " ~
r + B + CJ'"
is concave and
J 'f 2.5 ( 0 ) J <: I0/ is ( I ,

0 )
in agreement with experiment.
We see that the leptonic widths impose an additional constraint

to the potential, and that harmonic oscillator potentials may be
dangerous.
One problem about lepton width is that of the ~", the 3.77
resonance predicted long ago as being a D state, and recently dis-
covered.
In a purely central potential an L =2 state could not be pro-

duced in the non-relativistic approximation. However, two things
happen:
i) There are relativistic corrections to the Weisskopf-Van Royen

formula for the leptonic width which produce a contribution
re tv GD\'\st ... I 22." (0) (2.

where R2 is the radial wave function of the L =2 state; this
is not quite enough to account for the leptonic width of
0.37 keV of the ~".
420 A. MARTIN
ii) There 1S unavoidably a coupling between the L = 0 and L = 2

state via tensor forces. These tensor forces have been esti-
26)
mated by Jackson by fitting the P state levels; in this
way one can account for the leptonic width. It is not correct
to say that a dominantly D state cannot have a non-vanishing
L = 0 wave function at the origin as was stated recently in a
preprint which has been withdrawn.
The problem of higher states in the charmonium spectrum
We have already spoken about the 3.77 GeV tjJlI state which is
slightly above the charmed particle production threshold. The
description of this state as a pure quark-antiquark pair is still
reasonably good. For higher states it may be questionable in the
sense that a resonance above the charm threshold can be made of the
physical charmed and anticharmed particles. This is what is called
by De Rujula, Georgi and Glashow "molecular charmonium,,27) and
which, as pointed out by David Jackson, is not very different from
thinking that the p meson is an L = 1 pion-pion bound state. On
the other hand, it seems illogical to disregard completely the se-
quence of higher states produced by the quark-antiquark potential
which has been so successful in the description of the lower states.
h t h e eX1st1ng
W1t .. . 1 an L -- 0 J PC = 1
potent1a . pred 1cte
state 1S . d
around 4.14 GeV. This is a bit too high for the prominent peak at
E = 4.028 GeV which has been extensively used as a source of D's and
D*.
However, it seems to me that the most reasonable picture which
has already been proposed some time ago on a pure theoretical basis
.. 28).
by Dashen, Healy and Muz1n1ch 1S that the confined quark-
antiquark channel is coupled to the physical particle channel. De-
pending on the energy these channels are closed (and then their
contribution to the wave function is very small) or open. In this
way the two pictures are reconciled and the only risk is that there
NEW PARTICLES 421
might be a slight double counting. The existence of physical chan-

17 29)
nels is taken into account in the work of the Cornell group ,
and of the "naive quark group" in Orsay30). Then one can understand
that the mechanism may lower the energy of 3S triplet state to
4.028 GeV. In fact, the very complete calculation of the Cornell
31)
group takes also into account the effects of the closed physical
channels on the "stable" particles like J/ljJ and ljJ' and show that
these effects are important but that agreement with experiment can
be obtained by "renormalizing" the parameters such as the quark
masses.
Probably the greatest success of the model is the prediction

by the Cornell and Orsay groups of the dominance of the n*n* decay
of the 4.028 GeV state. If spin factors and phase space are re-
moved one finds that the n*n* cross-section is 100 times too large,
while the n*n + n*n cross-section is five times too large.
Le Yaouanc et ale show that this is due to the nodal structure of
the wave function in p space. They use harmonic oscillator wave
functions, which, as we have seen, is objectionable, but their
basic conclusion is certainly unaffected by this technical simpli-
fication.
5. THE NEXT SPECTROSCOPY
The arguments for a new heavy quark required by neutrino phy-

sics may have disappeared, but the discovery of the Lederman group
is there: a very clear bump in p + Be + ~+~- + anything has been
seen at 9.5 GeV with, if interpreted as a single state, a width of
1.2 GeV. Additional statistics are required to clarify the struc-
ture. Predictions have already been made by Eichten and Gottfried.
With a quark mass of 5 GeV we may have three states with a spacing
of about 400 MeV which fits with the observed width 32 ). A more de-
tailed study on this topic has been recently made by J. Ellis et
ale 33)
422 A. MARTIN
NOTE ADDED AFTER THE HAMBURG CONFERENCE
With additional statistics Lederman sees clearly two peaks at

9.4 and 10.1 GeV, and possibly a third one. If these two peaks are
made by the same quark-anti quark pair, they cannot be explained by
the same potential as charmonium because the spacing between the
levels is approximately equal to the ~~' spacing. Quigg and Rosner
notice that a potential Clog (r/ro) would produce this. Let us
notice that such a potential satisfies the conditions of our
Theorems I and II and is concave (Theorem III). What remains to
be established is whether it can admit a third bound state stable
with respect to the generalized 021 rule. Anyway, the third peak
is not clearly established.
NEW PARTICLES 423
REFERENCES
1) B.W. Wiik and G. Wolf, Electron-positron interactions, DESY

77/01 (1977), and Les Houches Summer School (1976).
2) v. Luth, SLAC-PUB-1873 (1977), Lectures at the Xth Internat.

School for High-Energy Physics, Baku, USSR (1976).
3) J.D. Jackson, Invited talk at the European Conf. on Particle

Physics, Budapest, 1977, to appear in the Conf. Proc. (ed.
L. Jenik), and CERN preprint TH.2351 (1977).
4) A.W. Litke, Invited talk at the European Conf. on Particle

Physics, Budapest, 1977.
G. Go 1dhaber, Invited talk at the European Conf. on Particle
Physics, Budapest, 1977.
U. Timm, Invited talk at the European Conf. on Particle Physics,
Budapest, 1977, and DESY 77/52 (1977).
5) V. Gupta and K. Koger1er, Phys. Letters 56B, 473 (1975).
6) T. Applequist et a1., Phys. Rev. Letters 34, 365 (1975).
7) E. Eichten et a1., Phys. Rev. Letters 34, 369 (1975).
8) L. Maiani" in "Elementary processes at high energy" (Academic

Press, New York, 1971), p. 610; Proc. 8 th Internat. School
of Subnuc1ear Physics, Erice (1970).
9) T. Applequist and D.H. Po1itzer, Phys. Rev. Letters 34, 43

(1975)
10) M. Chanowitz and F. Gilman, Phys. Letters 63B, 178 (1976).
11) H. Harari, Phys. Letters 64B, 469 (1976).
12) G. Karl, S. Meshkov and J.L. Rosner, Phys. Rev. D 12, 1203
(1976).
P .K. Kabir and A.J .G. Hey, Phys . Rev. D 13, 3161(1976).
H.B. Thacker and P. Hoyer, Nuclear Phys. B106, 147 (1976).
13) This results from a discussion between P.K. Kabir and the
author.
14) J.D. Bjorken, unpublished, quoted in Ref. 2.
15) H. Schopper, Lectures at this School.

424 A. MARTIN
16) B.W. Wiik, Lectures at this School.
17) E. Eichten et a1., Phys. Rev. Letters 36, 500 (1976).
18) A.B. Henriques et a1., Phys. Letters 64B, 85 (1976).
19) H.J. Schnitzer, Phys. Rev. D 12, 74 (1976).
20) R. Barbieri et a1., Nuclear Phys. BIOS, 125 (1976).
21) F.M. Renard, Phys. Letters 65B, 157 (1976).
22) A. Martin, Phys. Letters 67B, 330 (1977).
23) H. Grosse, Phys. Letters 68B, 343 (1977).
24) See, for instance: J.D. Jackson, Lectures at the SLAC Summer
Institute (1976).
25) A. Martin, Phys. Letters 70B, 192 (1977).
26) J.D. Jackson, private communication.
27) L.B. Okun and M. Vo1oshin, Z. Exsp. Teor. Fiz. 23, 369 (1976).
A. De Ruju1a et a1., Phys. Rev. Letters~, 317--(1977).
28) R.F. Dashen, J.B. Healey and I.J. Muzinich, Phys. Rev. D ~,
2773 (1976); Ann. Phys. 102, 1 (1976).
29) K. Lane and E. Eichten, Phys. Rev. Letters 12, 477 (1976).
30) A. Le Yaouanc et a1., Phys. Letters 71B, 397 (1977).
31) A more detailed account of the work of the Cornell group is

in preparation.
32) E. Eichten and K. Gottfried, Phys. Letters 66B, 286 (1977).
33) J. Ellis et a1., Nuclear Phys. B131, 285 (1977).

DIS C U S S ION
CHAIRMAN: Prof. A. Martin

Scientific Secretaries: A. de la Torre, D. Antreasyan
DISCUSSION No.1
LIPKIN:
TheX states should not decay into vector-pseudo-

scalar if the X~ are SU(3) singlets. The decay into
tensor-pseudoscalar (A2, K K) is allowed for 1+ and 2+
but forbidden for 0+. Are there data to check these
predictions and verify the quantum number assignments?
MARTIN:
I don't think that the data has been analyzed to

such detail.
SCHOPPER:
In the data that has been presented in the recent

conferences I can't find any of the final states that
you mentioned.
ROLANDI:
I would like to know why theX (3.45) has been seen

at DESY and not in the r
spectrum of the ~' decay at
SPEAR. Is it because of the trigger, or due to the
small branching ratio?
425
426 DISCUSSION
MARTIN:
I am not the best qualified person to answer the

question, but I think it is because of the trigger.
ZICHICHI:
At SPEAR they are rather insensitive to ~'s.
SCHOPPER:
I think that's true, but it is not fair to say that

theX (3.45) has been discovered first at DESY. The X
(3.50), the first intermediate state, was seen first. at
DESY, however the X (3.45) has been seen at about the same
time both at DESY and SPEAR.
BALDINI:
Supposing one has a potential that gives a satis-

factory fit to the charmonium data, what do you think
can be learned from the parameters of such a potential?
MARTIN:
First of all, if one likes QCD, one can use the

short range part of this potential to deduce the effect-
ive coupling constant of QCD. The other thing that can
be done, and has been done by the Cornell group, is to
take this potential as a universal interaction between
quarks and by inserting quark masses predict not only
charmonium states but also charmed meson masses. In this
way the Cornell group has predicted a mass*of 2.00 GeV
for the F and a mass of 2.15 GeV for the F , in excellent
agreement with experiment.
BIZZARRI:
I am very impressed to see how fast the L=l cc

states have been found, as compared to the L=l ss or uu,
dd states. This is largely due to the small width of
the ~' which allows detection of radiative decays, but
it is also due to the relatively clean e+.- initial
DISCUSSION 427
state. I think we have an important field here to be

studied by low energy e+e- machines (Adone, DCI).
A question that comes to mind is whether this potential
description of the cc states will be of any use towards
a better understanding of the lower mass states.
MARTIN:
I think one of the reasons why the potential descrip-

tion works so well is that the large mass of the charmed
quark allows us to make a non-relativistic approximation.
This approximation will not be so good for the low mass
quarks, and ~herefore predictions for the low mass states
will be more qualitative and less quantitative.
CELMASTER:
It should be noted that the D and F are also relativ-

istic. The reduced mass of the quark pair in the D and
F is only twice that of the n or p systems. Predictions
for the D mass including relativistic effects however
have been worse than those of the naive non-relativistic
models.
LIPKIN:
In response to Dr. Bizzarri, the reason that charm-

onium spectroscopy is better known than strangeonium is
signal to noise. The success of charmonium (and of the
new i (9.5 is due to the large electromagnetic couplings
to photons and lepton pairs - to be observed against a
very large hadronic background. The strangeonium sign-
ature, kaon pairs, has an appreciable background even
from non-strange states. The recent DESY experiment on
the F is an excellent example of the use of a low noise
signature to maximize the signal to noise ratio. Similar
ingenious signatures might be useful for strangeonium.
WETZEL:
I would like to ask if the calculation of the L=2

D-states of charmonium includes the annihilation channels
(such as DO) or uses only the cc potential.
428 DISCUSSION
MARTIN:
The first calculations only used the potential and

predicted the D-state correctly. When the Cornell group
included the annihilation channels some of the parameters
changed, for instance the mass of the charmed quark, but
the energy of the D-state mass remained more or less
unaffected.
BERLAD:
In my naive understanding an asymptotically free

theory should have a potential behaviour as shown in (A),
for example a harmonic oscillator potential flattened
out near the origin,
V<rj VCr)
r
A
and not the Coulomb-like potential (as in B) of the
charmonium calculations, where one obtains:
V(Y') rv
-r.. o
...1
T
This potential does not seem to lead to asymptotic free-

dom at short distances.
MARTIN:
It is not quite free. At short distances it should

behave like a weakened Coulomb potential, for example:
1.
The weakness of the potential will then be due to the

smallness of the effective coupling constant . at high
s
energies.
DISCUSSION 429
DISCUSSION No.2
WETZEL:
I wonder how strong the evidence is for a linearly

cc potential. Do you know of a calculation, including
complications like spin-spin interactions or the contrib-
ution from annihilation channels, where the power of the
rising potential has been introduced as a parameter to
be fitted to the data?
MARTIN:
I don't know of a calculation of spin-spin effects

other than for linear rising potentials. H. Grosse and
I have played numerically with central potentials of the
type _ A/'Y' + 8 y.oI with 2. ~ col." 2 \-Je found that as long
as one restricts oneself to the~~' and the p states,
potentials with 111' ot ~ ~h. give re~sonable agreement. If
one also includes the 4028 state as the second radial
excitation a power as high as <=to: 2. yields very severe
difficulties.
DINE:
You remarked that the 8 function which appears in

the spin-spin interaction can be controlled in atoms by
a cutoff of the order of the nuclear radius. My question
is about positronium. Does the Compton wavelength prov-
ide a cutoff, or does one simply accept the success of
the perturbation theory?
WIGHTMAN:
The matter of the effective cutoff in positronium

is tricky. In his Thesis R. Ferrel found the effective
non-relativistic Hamiltonian including~igher order eff-
ects. Several of the effective potentials were very
singular and it was not clear what the effective cutoff
should be. The matter was not completely cleared up
until the relativistic calculations of Karplus and Klein.
430 DISCUSSION
SCHECHTER:
This morning you commented on the surprising small-

ness of the electromagnetic mass splitting. Could you
elaborate on this?
MARTIN:
It is a general assumption that the mass difference

between two particles in the same isospin multiplet is
of electromagnetic origin. However we must recall that
QED does not provide a framework to calculate mass diff-
erences, the reason being that due to the renormalization
the masses are not calculated in the theory but are put
in it by hand.
CELMASTER:
There are gauge theories where the u-d mass differ-

ence is not input. It is calculable.
MARINESCU:
The latest data which you have shown regarding the

electromagnetic mass difference of the charmed mesons
indicate that the D+ - DO mass difference is smaller than
the D"_I>"o mass difference. This would be in contradict-
ion with the potential quark model.
MARTIN:
There is a contradiction between two sets of measure-

ments of the D+ and DO mass. Old results gave 1874 and
1865 whereas more recent numbers are 1868!3 and 1863.6*3.
So I don't think that the mass difference is well establ-
ished yet. In fact there also seems to be a contradict-
ion in the mass difference of D~o and D+. The values
given at the Budapest conference would not allow the
decay [)IfO... b't+ 'II
MOTTOLA:
In the phenomenological models like the MIT bag or

the potential models you have been discussing, where wave
functions and electromagnetic corrections are completely
DISCUSSION 431
calculable, there is no mystery in the mass differences

and everything is fully understood.
MARTIN:
Yes, at the phenomenological level there is no

problem.
MIETTINEN:
What is the most important problem in "old" spectro-

scopy which has been solved as a result of the new in-
sights gained in studying the charmonium spectroscopy?
MARTIN:
I can't give you a specific example. For me person-

ally, the most important result is that we have gained
much confidence in the idea that hadrons are bound states
of quarks and antiquarks.
MARINESCU:
I would like to give an example of a problem of the

old spectroscopy which has been solved after charmonium
spectroscopy was invented. It is the ~-A mass differ-
ence which can be explained by the spin-spin interaction
in a Coulomb-like one-gluon exchange potential.
BERLAD:
We were told that the potential model will be critic-

ally tested in the upsilon (9.5 GeV) region. Can this
model really be extrapolated to such high mass states,
and in case of disagreement with future experiment, won't
there be a too easy way out by just claiming, e.g., that
flavour invariance is broken?
MARTIN:
The calculation of the. (upsilon) may not be a very

firm prediction but it certainly can be attempted with
the insight gained from charmonium. The Cornell group
has done the calculations using the potential from
432 DISCUSSION
charmonium and in fact they obtain a triplet of states

with mass differences consistent with the interpretation
that the 9.5 GeV structure in Lederman's data, contains
more than one resonance.
MIETTINEN:
I think that the spectroscopy of the new super heavy

(SGeV) quarks that build up the upsilon will be a very
important testing ground of how much we really understand
about hadron spectroscopy. Somebody may think that re-
peating the charm story at a higher mass is boring. I
think just the opposite. We have now a new parameter to
vary: the quark mass.
LIPKIN:
The old spectroscopy has been bogged down by question

of the existence of the At and , ' . The charmonium analogs
of these states are clearly seen and also the electromag-
netic transitions between them. This gives us confidence
that the Rot and ~' probably exist even though they are
much harder to see.
BACE:
Could you outline the proof of the theorems you

mentioned this morning?
MARTIN:
I can sketch the proof of one of the theorems. Let

us say E 2.P ~ ['loS if (d.~Y')r. Y'l V, >O. The idea is to start
from the case where the levels are degenerate and switch
on an interaction depending on a parameter A.
v=-- ~.,. 2. + A2. V, (A I Y')

for 1\ = 0 this reduces to the Coulomb interaction for
which we have degeneracy. For small A we can apply per-
turbation theory
E' l S (,\) - E 1 pC,) ) ~ Xl. JV, c" r ) (Lll_ V 1.) dr

DISCUSSION 433
where lA. is the 2S wave function and V the 2P wave funct-

ion. Now," has one node and v two nodes. It is easy
. h es only twice.
to prove t h at U. ~ - v ~ van~s Now we have
the extra constraints
= o (normalization)
j U; v'&. elY": 0 because
a pure Coulomb perturbation does not remove the degener-

acy. So
El~_E2P ~,\1 ffU~\Tl)[r\.{(Ar)+A+BtJ oLr
where A and B are arbitrary. So if TV, is convex, we can

choose A and B such that the bracket changes sign where
U~ v'changes sign and prove that the integrand is always
positive. r ~ convex follows from our stronger condition.
To avoid perturbation theory one continues inA and
tries to see if E.2.sU\) - E 21' ei\ ) can vanish. At the point
where it would vanish the derivative
:~ [E2S (it) - E2.p (1\)]

is proportional to
f (" 2 - \T 1 ) [2 V, (Ar) .. (A r) ~ I ( Ar) ] d V'

but we have additional constraints
j(ut_\T~) elY' =0
and from the virial theorem
~(ut_v1.1 (2V + r dr =0 , 1:)

where V is the total potential. By combining these
quantities and using the conditions of theorem I it is
possible to show that
~(EH-t1P)
lA).
is positive. This is impossible. Therefore IS - 1 P
remains always positive.
NEW PARTICLE PRODUCTION IN lUillRONIC INTERACTIONS
Hin Chen
l1assachusetts Institute of Technology
Cambridge, Mass. USA
EP Division, CEP~, 1211 Geneva 23, Switzerland
INTRODUCTION
For many years the fundamental building blocks of the universe

were thought to comprise only three types of quarks plus 4 types of
leptons. In 1974, a new particle J(l) with 3 times the mass of the
proton and 1000 times longer life time than an otherwise similar
vector meson, e.g. the p meson, was discovered in proton-nucleus
interactions. The J particle is currently being interpreted as
the bound state of a charmed quark anti-quark system. To understand
the nature of the charmed quarks and to find out how many other new
quarks there are, an enormous experimental effort has been expended
to study the decay and the production of the J particle with various
beams on r.lany different targets, and also to search for more J like
particles to much higher masses. Hadronic production is a powerful
tool for the following reasons :
1) the cm energy ~ available in a proton accelerator ~s higher

than in other types of accelerators.
2) The intense hadron beams available allow high sensitivity to

be achieved in the particle search.
435
436 M. CHEN
3) One can vary the type of beams and targets in the experiments
to have a better understanding of the nature of the production
mechanism.
4) The mass spectrum in the final state covers wide range and
therefore is suitable for searching for new particles with
unknown masses.
5) One can study the correlations between the new particle with
the rest hadrons.
Today's topics, concerning the recent experimental study of

new particles in hadronic reactions, are listed below :
I A. The energy dependence of the J production cross-section.

B. The dependence of the J production cross-section on X, PT ,
atomic number of the target and helicity decay angle.
c. The relative production cross-sections of the J with
+ + -
IT-, K-, P and P beams.
II Search for charmed particles in the dihadron final states,

with and without additional lepton trigger.
III Recent discovery of a new massive resonance in dimuon final

state.
IV Dimuon physics at ISR.
I A. Energy dependence of the inclusive J production cross-section

in the reaction
P+ nucleus +J + anything
+ - +
4. e + e or)l +)l
As shown in Fig. 1 the proton induced inclusive J production cross-

.
sect10n per nucleon do I . the rap1d1ty
where y 1S . . def1ned
. by,
dy y = 0
NEW PARTICLE PRODUCTION IN HADRONIC INTERACTIONS 437
oII
>-
x Cord en et 01.
bl>- Kourkoumelis et 01.
"C"C
t o Amoldi et at.
181 Busser et 01.
o Snyder et 01.
Anderson et 01.
<> Antipov et 01.
l;. Aubert et 01.
VS in GeV
Fig. 1: s dependence of the p-induced J production cross section.

438 M.CHEN
y = 1ln E+pP L , with E = the energy, PL = longitudinal momentum of

E- L
the J, ~s found to increase very steeply with energy
i) Near thresho ld dOl

~ .
~ncreases b y an order 0f '
magn~tude
y y = 0 (2)
from IS = 6.S to 7.S GeV as measured at BNL
ii) from BNL to FNAL energy range the cross-section increases
by another order of magnitude (3)
iii) At the lowest ISR energies IS = 23.S and IS = 30.6 GeV the
cross-sections are in good agreement with the FNAL measure-
ments at comparable c.m. energies. Even at such high energies
the J production cross-section continues to increase with
(4)
energy ,although, at a somewhat slower rate :
do
dy y = 0
Vs 62.4)
5.S
+ 5.8
- 1.8
do (IS 23.4)
dy y = 0
To conclude, we note the cross-section has increased by three

orders of magnitude from IS 6.5 to 62.4 GeV, steeper than the
production cross-section of any other particle measured so far.
One possible mechanism which would enhance the J production at

higher energies is the associated production(S) of the J with a
pair of charmed meson e.g. DD.
- At lower energies this process is
kinematically suppressed, whereas at higher energies it may become
(6)
more and more dominant. Two groups at FNAL have searched for
evidence of the reaction
p + A -+ J + D + D + X
L
~
f.l
+
f.l l- f.l + X'
and have se t an upper limit
o( J DD )
~ 0.0 1
o( J )
where a branching ratio of D + ~ + X~ 0.1 was assumed~7) These

experir.lental results show that up to IS ~ 25 GeV, the associated
production of the J with D charmed mesons is negligible. In the
near future, several ISR experiments will be able to test the
importance of this mechanism at IS ~ 62 GeV.
I B. The dependence of the J production cross-section on X, PTi

atomic number of the target and helicity decay angle :
Host of the work on these topics was carried out by the

Chicago-Princeton group at FNAL. Since Prof. A.J. Smith is going
to present these data himself tomorrow, I will just outline the
results here and refer you to his lecture in this same proceedings
for more details.
The variable X is defined as ~ = 2 PL* / IS. For J prod-

uction, the cross-section are commonly parametrized as follows :
E A e - B PT (1 - x ) n
dxdp 2
t
For proton induced reaction, one finds
B 2.05 0.09
n 3.44 0.14
+
while for n-one obtains
B 2.0 0.1
n 1.5 0.2
The interesting feature is the X dependence of the ratio of the P

over the n induced cross-section is (1-X)2 which was explained
nicely by quark counting in the quark-antiquark annihilation
model(8)
440 M,CHEN
In hadronic reactions, of the ordinary hadrons such as

T
n, K, P etc.is around 350 }IeV. One of the surprising observations
of the J particle production in hadronic reactions is that the
'
measure d 0 f t h e J ~s " f '~cant 1
s~gn~ y greater t h an 350 MeV (9) an d
T
is independent of s. Furthermore when is plotted against the
mass of dimuons in the region between the p meson and the J meson
(Fig. 2), it is found that the increases with mass from
Similarly the of dihadrons was also
""500 MeV to ""1000 }leV.
observed to increase with the pair mass of the hadrons in the
region 2 <m < 5 GeV at IS = 7.5 GeV.
The A dependence of the J cross-section on nucleus with atomic

number A is usually fitted to a form Aa . The data show a is slight-
ly less than 1. a = 1 implies that the incident beam particle can
penetrate the nucleus freely and interact with any nucleon inside
the nucleus independent of all the other nucleons, or in other words,
the absorption cross-section is negligible and each nucleon serves
as an independent source of the dimuons. In figure 3 the measured
values of a versus the dimuon pair mass is shown not only including
the J but also the p, w, and the continuum. We notice an inter-
esting trend : as the pair mass of the dimuon increases, the absorpt-
~on cross-section decreases and a approaches 1. In quark pictures,
the higher the mass of the dimuon, the more energetic 'the quarks
will have to be. So the data is telling us something about the
interaction of the quarks as function of their energies.
In the naive quark antiquark annihilation model, since the

helicity of the quark (antiquark) is -1 (+1), one expects the
resultant photon has helicity -1 and therefore is polarized. The
actual distribution of the lepton pair therefore has the form
2
1 + P cos e with P = polarization to be slightly less than 1 due
to the transverse motion of the quarks and some final state inter-
actions. Experimentally it was found P = -0.28 0.22 consistent
pC
--..-..... . -+--+--+-
2
M (GeV/c 2)
Fig. 2: Plots of versus mass for TI+ TI- and protons incident
on carbon. T
442 M.CHEN
I I I I I I
a
+
J -
1.0 f-
For p induced : ---T--- .J..
r
- - --.---+---+- -
p_w
0.5 - -
++
/ V
,II' I'I'
/
1.0 - e"." '"'000' , J -
--+--+
-*----T--- 1
T
-
p_w
0.5 - -
I I I I I
0 2 3
M (GeV/c 2 )
Fig. 3: Atomic number dependence of muon pair production cross

section versus mass.
with zero. This polarization measurement should be carried out at

higher energies for higher mass dileptons, for the following reasons:
1) It is an important test of the Drell-Yan type quark-antiquark

annihilation model.
2) When the weak intermediate vector boson Z is produced some

0
day in p-p interaction, if the Z is polarized the unique
0
feature of such a weak interacting particle 1S to measure the
charge asymmetry in the distribution of the momenta of the
leptons and anti-leptons. Therefore it is important to know
if a non-vanishing polarization can be expected.
I C. Production of the J particle with different incident beams :
After the discovery of the J particle, there are many specul-

ations about its production mechanism in hadronic reactions: In the
Drell-Yan(ll) type quark-antiquark annihilation models ordinary
quarks (mainly from the valence) and anti-quarks (from the sea for
nucleons and from valence for n, K and p) form the J via a coupling
which is experimentally known to be suppressed. Since there is much
more ordinary antiquark component inside an antiproton than inside
a proton, this model predicts the proton induced cross-section
a(p) to be Quch smaller than the antiproton induced cross-section
a(p), especially near the threshold where valence quarks dominate

over sea quarks. In the charmed quark-antiquark annihilation
model (12) both charmed quark and antiquark come from the sea to
form the J via a strong coupling constant. In the gluon amalgamation
model (13), gluons from the hadrons scattered to form some P-wave
intermediate state such as X (C = + 1) which subsequently decays
into the J plus a photon or vector meson. Since we expect equal

amount of the gluons or the charm quarks from the seas of any of
+ +
the ordinary hadrons (n-, K-,p or p), the last two models predict
+ +
the J production cross-section to be the same for n-, K-,p and p.
444 M.CHEN
The relative particle/anti-particle production cross-sections

(14)
of J were measured by the C-M-N-E-R group. The production of
the J decaying into ~+~- was measured at 39.5 GeV/c using both
negative and positive unseparated beams from the CERN SPS incident
(15)
on a copper target located in the Omega spectrometer Product-
~on of J was observed with all six beam particles (n, K, p and p)
and relative cross-sections have been obtained for J production
for X > O.
The apparatus is shown schematically in fig. 4. Three threshold

Cerenkov counters were used to identify incident beam particles.
The target consisted of five 4 cm thick copper slabs interleaved
with scintillation counter Tl-T6. A 1.46 m thick copper obsorber
immediately after the target reduced the hadron flux by a factor of
thirty. Particles emerging from the absorber were detected with a
four-element scintillation counter hodoscope, S6. Particle traject-
ories were recorded with a TV readout of forty optical spark chamber
gaps, in a region of 1.7T average magnetic induction. Particles
leaving the chamber system passed into the 1.25 m thick iron return
yoke of the magnet which acted as a second hadron obsorber before
reaching the last counter hodoscope. The observed J width ~s
2
consistent with the 110nte Carlo estimate of 0.35 GeV/c .
+ - +
The ~ ~ @ass spectra produced by nand n beams shown ~n
2
Fig. 5 exhibit clear J signals with masses 3.12 and 3.15 GeV/c
and FWHM's 0.43 GeV/c 2 and 0.36 GeV/c 2 respectively.
The mass distributions for ~ +~- pairs produced by K , K+,

P and p are also shown in Fig. 5. All exhibit clear enhancements
at the J mass. Strong signals from p/w + ~~ and, for K beams
only, shoulders from +~~ signals can also be seen.
The results for J production on copper at 39.5 GeV/c beam

momentum for X > 0 are
Hodoscope
IEuIEuIE~ EulEull
I ~ t I
1 12 l T4 T5 T6
Spark chombt>r
Torgl't modutes
sc:int illotors
Beam
SI
Fig . 4
N N
V V
~ :><II
I.!) I.!)
N <'j
d 0
VI
$
c 1:
<II
<II
> >
<II <II
Q. ::.:::
tei Tt Tt' I::
I
p
0.1
n
0 .21 0.50 1 2 3 5 0.21 050 1 2 3 5
Fig . 5
446 M.CHEN
a (n ) a (K) : a (p) = 1 1. O. 0.3 : 1. 0* 0.3

The particlel antiparticle ratios are again for X> 0 :
a (n+) I a(n-) 0.87 0.14,

+
a(K ) I a(K ) 0.85 0.50 and
a(p) I a(p) 0.15 0.08.
The conclusion is the data are consistent with quark annihilation

model while the gluon amalgamation and the charmed quark fusion
model which predict each ratio to be unity are in clear disagreement
with the pip ratio.
II Particle Search in Hadronic Final States
If the J particle is the bound state of the charm quark and

anti-quark, we expect many particles made of one ordinary quark and
one charmed quark to be also produced in hadronic interactions.
Since the ground state charmed particles are all stable against
strong decays and therefore are expected to have relatively long
lifetimes ( T <10-13 sec ), the most direct means of detecting them
is to observe a narrow peak in the effective mass spectrum of
several hadrons.
(a) Hadron Pair Final States
The BNL/HIT group (16) has observed oppositely charged hadron

pairs in proton-nucleus collisions with good particle identification
and good mass resolution ( a ~ 5 HeV). Laboratory angles of the
spectrometers pictured in Figure 6 were chosen so that particles
were observed at ~ 90 0 in the center-of-mass system and hence the
two-body hadron system was produced near X~O in the center of mass.
To avoid systematic errors, 7 overlapping magnet settings were made
for the measurement.
C 5
_ ~ ~\\\\Y
Beam I
/1
Target
Plan view of the detector
b 2m
Mo' HI ' M2 - dipole mngncts
Ao ' A, B, C - 8000 proportional wir chambers
a , h - 8x8 hodoscop s
S - 3 banks of pb-glass shower counters
CB, Co Ce - gas Ccrenkov count rs
Beam
~
Target
o 2m
Fig. 6
448 M.CHEN
Figure 7 shows the results of some of these typical reactions.

\lithout acceptance corrections the yield increases with mass due to
an increase in acceptance. It then decreases due to the decline in
production cross-sections. There are no sharp narrow resonances in
any of the 9 reactions. The upper limits corresponding to a Sa
signal are listed in the following table :
, "
Sens~t~v~ty
(cm)
2
For Narrow Resonances
Ah x
7T
+ -
K
2.25 GeV
1 x 10- 33
3.1 GeV
4 x 10- 35
3.7 GeV
1 x 10-35
+ -
KIT 4 x 10-33 8 x 10- 35 4 x 10-35
p P
- - 4 x 10-34 2 x 10-35
+ -
KK 1 x 10-33 5 x 10- 35 1 x 10-35
+ -
7T7T 8 x 10-33 5 x 10-34 3 x 10- 35
'h'~gan-Pur due group (17) has searched for

Th e FNAL-M~c product~on
,
of heavy particles decaying into all two-body states h h% with

charge + 2, 0, or -2 in proton-nucleus collisions at 400 GeV/c.
The experiment has mass resolution of 15 MeV FWHM obtained by
.
(.)
......
>Q>
~
\0
C\i
......
~
C
Q>
~
0
eo
40
Fig . 7
450 M.CHEN
using a drift chamber system. Good particle identification 1S

obtained in each arm of the spectrometer over a momentum range of
7-21 GeV/c with three threshold Cerenkov counters. Present cross-
-30 2
section sensitivities at l\.IT 'V 1.86 are 0DBKIT < 0.8 x 10 cm.
At 400 GeV/c, OJBWW'V 10 nb.
. "
A slm1lar search was made at t h e ISR 1n
" p-p co 11""
1Slons (18) at
IS 'V 52.7 in neutral hadron pair combinations where PT 'V 0.7 GeV/c
and x 'V 0.3 for the pair. The limits quoted for ~IT = 2.0 are
+
d 30 x 10- 30 for K IT
E .B <
6.
+
dp3 2.1 x 10- 30 for K IT
P .73 GeV/c
T
x .32
\fuereas the dihadron spectra do not show any sharp resonance

states, the cross-sections ~~ vs TIl for groups A (IT p), B (IT+IT-,
+- - +- +- + -+ ""
pp, KIT, K p) and C (K K , K K , IT p, KIT) do exh1b1t some simple
degeneracies.
The MIT-BNL group observed that at IS =

7.4 GeV, the cross-
-5m
sections for each group decrease with mass 'V e and differ from
" (16)
each other by an order of magn1tude . Similarly, at Is = 25 GeV,
-4m
the F-M-P group found oem) Q e The average PT of the dihadrons
" "
was found to be 1ncreas1ng "h t h
W1t " mass 0 f thedOh
e pa1r l a d rons (16)
similar to the P~ behaviour of the dimuons which we discussed
.1
earlier.
+ -
(b) In the Wh h mode
+ -
As we discussed in (a) charm was not observed in the h h mode.
There are however two possible explanations that charmed particles
may be produced abundantly without giving significant signal in the
+ -
h h mode:
1) The branching ratio of two-body decay mode may be very low;

+ -
2) The noncharmed h h cross-sections are much larger than the
charmed ones such that the latter is swamped.
In order to enhance the signal due to charm, the following

observations were made :
1) In hadron interactions charmed particles are produced in pairs

in order to conserve quantum number.
2) There are theoretical speculations that the leptonic or semi-
leptonic decay modes are of substantial size (8%).
Therefore an event selection for h+h- with a ~ as a fragment

from the other charm partner detected will be a much more selective
signature of charm. One such possible reaction is
l Lu
-
p + P +D+ D + x
or Kp if Baryon
KW or A ~\i
The BNL/~1IT group(19) added a collimator of tungsten, uranium, lead,

Fe, and concrete near the target area of the double arm spectro-
meter. Muons were identified by pulse height in scintillators
shielded from the target by the collimator and by time-of-flight
between these counters and the spectrometer hadron pair signal.
A muon signal for P~ > 0.7 GeV/c has been clearly demonstrated.
- +
No significant structure was observed in the K 1T mass spectrum
produced in association with a ~ and an upper limit of
2
cm
452 M.CHEN
was . d(20) 3 x 10- 3 , one gets

obta~ne If B~BKIT ~ (0.1) (0.03)
A similar experiment is carried out in 400 GeV/c proton-nucleus

collisions using a modified version of the FNAL-Mich-Purdue appara-
tus at Fermilab. Preliminary results show the upper limit on aD
with ~ trigger is again about the same as that without muon trigger.
III Observation of a Dinruon Resonance at 9.5 GeV in the dimuon

final State
A strong enhancement is observed at 9.5 GeV mass in the final

state of dimuon by the C-F-SB group(20) in GeV proton-nucleus
collisions. The produced muons are analyzed in a double-arm mag-
netic spectrometer system with a mass resolution ~m/m (r.m.s) ~ 2%.
The spectrometer is shown in Fig. 8. Narrow targets (~.7 mm)

with lengths corresponding to 30% of an interaction length are
employed. Beryllium (18 interaction lengths) is used as a hadron
filter, covering the 50-95 mr (70-110 0 CM) horizontal and 10 mr
vertical aperture in each arm. Polyethylene (1.5m) and a 2.2 m steel
collimator complete the shielding.
The spectrometer dipole magnets deflect particles vertically,

decoupling the production angle of each muon from its momentum
determination. At full excitation (1500 A), the magnets provide a
transverse momentum kick Pt ~ 1.2 GeV. Conventional PWC's and
scintillation hodoscopes serve to define the muon trajectory down-
stream of the air dipole. Following the PWC's is a solid iron mag-
net (1.8 m long, energized to 20 kG) used to partially refocus
the muons vertically and redetermine the muon momentum to 15%.
A threshold Cerenkov counter on each arm also helps prevent possible
low momentum muon triggers. The apparatus is arranged symmetrically
453
NEW PARTICLE PRODUCTION IN HADRONIC INTERACTIONS
P ll
PJO
P5
--
--r-
--
95
t -_____ ~r~_~~~____ l_l __

MRAD
1.5
~ Steel
III Hevimet I$cint
eTR
fIT] Beryllium
1.0
0 .5 O eH ,
~ pwe
o 25 30
Meters
Fig. 8. Double arm spectrometer at FNAL for the discovery of the

new resonance.
454 M.CHEN
with respect to the horizontal median plane in order to detect

+
both ~ and ~ in each arm.
16
The dimuon data represents 3 x 10 proton at 400 GeV inter-
acting with Cu or Pt targets. The acceptance of the detector is
IV 7 x 10- 3 for - 0.3 <y < 0.3. The measured ~Tidth of the J
particle is about 200 MeV (FWHM) which serves as a calibration.
The dimuon mass spectrum is shown in Fig. 10. The spectrum,

apart from the region of the enhancement, appears to be a smooth
-Bl>l
continuum which has a mass dependence as Ae where M is the
dimuon mass. In the region 8.8 < H < 10.6 GeV, the data contains
a factor of 2 more events than what one would expect by inter-
polating from the continuum. Fitting the surplus events with a
simple gaussian of variable width yields the following results :
Mass = 9.54 0.04 GeV
do -37 2
B/ 3.4 0.3 x 10 cm /GeV/nucleon
dy
y = 0
FWHU= 1.16 0.09 GeV
X2 = 52 with 27 degrees of freedom.
Since the ratio of X2 over degrees of freedom is large, the

result is statistically unlikely. Since the observed width of the
enhancement is greater than the apparatus resolution of FWHM IV 0.5
GeV, it is possible there are more than one resonances close
together. Fitting the surplus events with two peaks where widths
are fixed at the resolution of the apparatus yields
11ass 9.41 0.02 GeV
do -37 2
B. dy I = 1.86 0.1 x 10 cm /GeV/nucleon
y o
l- I I I
I- -
+
I-
++ +++ +
I-
++ +
I-
++++++ +
I- -
++++
-
l-
t
o +H
t
II
>-
71- -
-
Nbl~
E
I-
I- -
"0"0 I-
tttt
I- -
I- -
I- -
f-
l-
tttt
I I I I
9 10 11
mlJ.lJ. (GeV)
Fig. 9. Dimuon mass spectrum measured with 400 GeV proton beam
at FNAL.
456 M.CHEN
and Mass = 10.06 0.04 GeV
-37 2
B. dO, = 0.74 0.07 x 10 cm /GeV/nucleon
dy
y o
2
with X 18.7 for 18 degrees of freedom.
The two peak hypothesis is statistically more favourable over

the single peak one. Also a resonance with one GeV width would
require that the hadronic decays dominate the leptonic decays,
leading to a production cross-section of the resonance comparable
to that of the J which is unlikely for such high mass. Another
interesting feature is the spacing between the two resonances.
I
0.6 GeV, the same as that of ~ and J.
In the same spirit as the J particle is interpreted as the

bound state of the charm quark-antiquark system, the new resonance
is most likely to be a bound state of some new quark antiquark
system.
IV Dimuon Physics at ISR
There is one experiment in progress at ISR, which I am person-

ally involved, looking for massive muon pairs in p-p collision.
With IS = 62 GeV. the ISR is the highest energy machine at the
present.
The Physics
In this experiment the detector is designed to study the

reactions :
(1) p + p +
Yv + X
L j/ \1-
where y is a time-like massive virtual photon, by measuring the

v
+
mass and angular distribution of the heavy photon y + \l \l in the
v
mass region 3 < m < 25 GeV. Information on the structure of X will
also be obtained.
(2)
up to ZO mass of 50 GeV, where ZO is the neutral heavy boson

responsible for the existence of leptonic and hadronic weak neutral
currents.
(3) Looking for more J-like new particles, ~.e. vector particles
with new quantum numbers other than charm. We believe there are
more such particles in the mass range 10 to 50 GeV region. Vector
particles, similar to photons, decay into lepton pairs.
(4) Checking scaling behaviour of the dimuon mass spectrum.

According to quark models, the dimensionless quantity m3 :~
m
should be only a function of ~ where m is the pair muon mass
and rs the CU energy of the proton beams.
(5) Looking into the behaviour of the hadrons produced together

with the massive muon pair. Do they have a dip in the distribution
because a large amount of momentum - energy is taken away by the
dimuon? Are there some unusual particles such as quarks produced ?
If quarks were ever going to be freed, here should be a good place
to look for them.
Figure 10 shows the set-up of the muon detector : muons are

detected and analyzed by iron toroids with uniform circular field
of 17 kG. Drift chambers sandwiched between the toroids identify
hadrons and large-angle Coulomb scattered muon pairs from hadron
decays.
.....
0'1
~
~-- .....
-..
p + P ZO + x -..
L,u fl-
~E Ulal ~
I'OKE YOKE
magneti zed , ,
qz,ozzTL1~"'-".~""'&
P
LUZ7//7Z7ZZZ7771lL21
.~~~~
., " ,,
", ,
----:
I s:
" I
,, I (")
I '- :I:
m
Z
Fig . 10 . Dimuon detect or at the ISR . niv . 441.60
The unique features of this detector are as follo\ls :
1) It covers an angular region 15 0 < e < noo which enables us

+
to detect very forward ~ ~ pairs. In the quark antiquark
annihilation model, the quark is normally more energetic than the
antiquark. As a result somey ~s are produced with large PL and
therefore decay into muon pairs in the forward cone.
2) The large solid angle = 8.4 sr enables us to detect Zo up

~Q
to mass of 50 GeV. The maximum detection angle e = 1100was chosen

In order to permit installation of a suitable detector to study
the recoil system x; further studies might well increase the
value somewhat.
3) The small opening angle detector enables us to detect 10 >

m > 3 GeV with high statistics and measure longitudinal (P Lo ) and
transverse (P.L) momentum dependence of virtual photons and new
vector particles, and thus study the electromagnetic properties of
protons in the time-like region at small distances.
+
4) since we detect a Yv + ~ ~ in the forward region, momentum
balance implies that the residual hadron X in reaction (1) would
favour the backward direction, and information on it is obtained
by the counters and drift chambers situated in the noo < e < 180 0
region.
In the selection of dimuon events, cosmic rays are rejected by

comparing the arrival time of the particles at the top and the
bottom counters which is separated by a distance of 5 m. The
arrival time is determined to 1.ns by measuring the mean time of
the two photo-tubes located at the ends of the counter. For pairs
of particles produced from the beam, the net difference in timing
should be zero while that for the cosmic ray is 15 ns. Further-
more when a cosmic ray muon goes through the detector to simulate
a pair, the two tracks are exactly back to back. For a genuine pair
even when the parent particle is produced at rest in the CM system,
460 M.CHEN
the two muons will be up to 10 0 different from back to back because

the ISR beams cross at 150 and thus has a center of mass motion of
S '" 0.12.
The event rate is estimated by scaling the measured cross-

section at lower energies to Is = 62 GeV. Fig. 11 shows after one
year of running the expected event rate. The calculated background
from hadron punch through the iron magnet and from K, TI decaying
into muons is negligible.
The detector is just completed now. A typical dimuon event is

shown in Fig. 12. We notice that most dimuon events are very clean,
seldom contaminated by spur ions tracks. This shows the shielding
power of the magnetized iron against hadrons. We are just taking
data now. How many more J-like new particles and ZO's nature has
stored for us in this mass range is anybody's guess.
10000
VS =62
.. VS =52
1000 -
u
----
:>
0.> NU
'-" 100
'-
0.> ----
:>
0.>
0- '-"
....,VI '-
0.>
c: 0-
0.>
>
UJ ....,VI
10
"""-
C
0.>
>
LLJ
"-
""
1
6 8 10 12 14 16 18
"20 1
20 25 30 35
m"" (GeV/c 2) mzo (GeV /c 2 )
a b
Fig. 11. Expected dimuon events from Yv (a) and from ZO (b).
oDD.
RUN ND . 715 EVENT 176
" ASS-16.S39 0.165 "fAP. 2.695 xr 0.575
p\- to . 'll! Xl- \.579 P2 --IQ.617 X2- 5 . 971,1
lOO.
MUON
......
200. iron yoke

~~ll.5.5
I
L,:::.dFl. h
100.
1.4
o. I IIU e
,100 .
2DO.
)00 .
Fig. 12. Typical dimuon event observed at ISR.

462 M.CHEN
REFERENCES
1) J.J. Aubert et a1. Phys. Rev. Letters 33, 1404 (1974). See also
e+e- Physics, same proceedings, by H.F. Schopper.
2) 11.Chen, Lecture at International Summer Institute in Cargese
on "Weak and Electromagnetic interactions at High Energies"
published by Plenum Press, 1976, Editor N. Levy.
3) B.C. Brown et a1., FNAL preprint (1977)
K.J. Anderson et a1., Phys. Rev. Lett. 36, 237 (1976)
B. Knapp et a1., Phys. Rev. Lett. 34, 1044 (1975)
4) C. Kourkoume1is, CERN 77-06 (1977)
5) S. Ellis et a1., FNAL preprint 76/29 - THY
T.K. Gaisser et a1., Phys. Rev. D13, 171 (1976)
6) 11. Binkley et a1., Phys. Rev. Letters 37, 578, (1976)
7) J.G. Branson et a1., Phys. Rev. Letters 38, 580, (1977)
P. Piroue, B.N.L. preprint (1976)
8) V.A. Matveev et a1., Lett. a1 Nuovo Cimento 7, 719 (1973)
S.J. Brodsky et a1., Phys. Rev. Letters 31, 1153 (1975)
9) J.G. Branson et a1., Phys. Rev. Lett. 38, 1334 (1977)
J.J. Aubert et a1., Phys. Rev. Lett. ~,416 (1975)
10) G. Fanar, Phys. Lett. 56 B, 185 (1975)
11) S.D. Dre11 et a1., Phys. Rev. Letters 25, 316 (1970) also
Ann Phys. (N.Y.) 66, 578 (1971)
12) A. Donnachie et a1., Nuc1. Phys. Bl12, 233, (1976)
H. Fritzsch, Ca1t 68-582, (1977)
13) S.D. Ellis et a1., Phys. Rev. Letters 36, 1263, (1976)
14) U.J. Corden et al., to be published
15) F. Bourgeois et a1., CERN/DD/DH/70-1
16) J.J. Aubert et a1., Phys. Rev. Lett. 35, 416, (1975)
17) D. Bintinger et a1., Phys. Rev. Letters 37, 732 (1976)
18) U.G. A1brow et a1., Nuc1. Phys. Bl14, 365, (1976)
A more recent upper limit from ISR (R606) sets BO' < 9}JP.
see ISR workshop/2-8/1977, Editor U. Jacob
19) M. Chen. Proceedings of the Vanderbilt Conference on New Results
in High Energy Physics,_ Editor R.S. Panvini (A.loP., N.Y., 1976)
20) S.W. Herb et a1., Phys. Rev. Lett. 39, 252 (1977)
DIS C U S S ION
CHAIRMAN: Prof. M. Chen
Scientific Secretary: S. Olsen
DISCUSSIon
SANDA:
_ I have 2 questions on J Itl' production by K.:t, TT:C1 P

and p beams. a) A naive reasoning indicates ~(K~ ( u{Kj
and (liP) / ~ (p) !lI u{Kfo) I u ( I< -) Can you comment on the
number ~(I<t)lrr(l<-)c. 0.85:f:.5 and on future prospects
for measuring this? b) You did not give a number for
~l~)/u(p), It should be available from the figures
you showed. Can you give it?
CHEN:
From the data of the 5L experiment one gets

u-(p)/v-(rr) = 0.14. On the first question, maybe
Prof. Smith can make some comments.
SMITH:
We will have 2000 K+ and K induced events in our

Fermilab experiment. The data on the u(~)/or{~)
ratio obtained by our experiment is much better than that
presented here and will be shown tomorrow.
JUNG:
I was on the n.
experiment. On the number u-{J<+)/rr(lt)
= 0.85 Z. 0.5, the error includes the statistical error
and a systematic error of 10%, which we think to be
463
464 DISCUSSION
realistic. We obtained the number of J/f. both with

a background subtraction and without, taking all of the
events in the J1r region. The difference is also in-
cluded in the error.
CHEN:
In addition to all these, for proton production of

the J there are also some corrections due to the prod-
uction of J's by secondaries produced in the target.
Also some J's can be reabsorbed in the nucleus, about
a 16% effect in all reactions.
MIETTINEN:
What model is used to generate the curves shown on

the graphs of the production ratios?
JUNG:
Drell Yan without color.
CHEN:
With or without color actually should not change

the ratios.
BACE:
Is the production on hydrogen or a heavy nucleus?
CHEN:
A copper target was used in theSt experiment.
BACE:
Apparently coherence effects can make a large

difference for production with protons. Another question.
Do these results exclude color?
DISCUSSION 465
JUNG:
The results are very sensitive to the quark
distributions used.
CHEN:
The data points are between the two hypotheses,
with and without color.
OLSEN:
In the graph you showed for the cross-section for
pp -+ JX, the cross-section at 40 GeV was 4 or 5 times
the cross-section at 30 GeV. This corresponds to a
change in CM energy of only 1.5 GeV out of 8. In going
from pp to trp, to pp the amount of available CM energy
changes by a comparable amount. Does it make any sense
to' compare the J production cross-section for differ-
ent incident particles at the same laboratory momentum,
when the cross-section is changing so rapidly?
CHEN:
I agree there is some problem. I think that at
least the cross-sections for pp ~ JX should be plotted
vs rs -2 'l"4J> to account for the limited available
energy, as was already indicated on the excitation curve
I presented.
PREPARATA:
It is not clear whether is - 2 ""'p is the relevant

variable for pp and (:r relevant for pp. J production
may not come from pp annihilation events.
CHEN:
In the charm quark or-Bluon fusion model I would
guess it should also be ~ - 2Np for the p case since
not all of the quarks are annihilated. Therefore the
ratio and the conclusion that it cannot be purely done
to these two processes remain to be valid. In the quark
antiquark annihilation model of Drell-Yan, it is not
clear which variable should be used. In the very extreme
466 DISCUSSION
(and unlikely) case that we use is for p and 6 -2 IMp

(j (Ii) / u (p) may reduce from
for p beams then the ratio
7 to 2 or 3, using the cross sections at ~ =,
GeV on the excitation curve for proton production of
and II
the J. This is unlikely since we do not see a peak at

X = I for p data, which is necessary if all quarks are
annihilated.
MINGUZZI-RANZI:
It is not clear to me what remains in the target

after two quarks annihilate for J production. In part-
icular when the target is a quasi free nucleon as in a
light target. Are there any results on this?
CHEN:
These are beam dump measurements and can't see any-

thing but /"''S. In the Fermilab Streamer Chamber experi-
ment, J events looked very much like other events.
WHITE:
I am confused by your use of the name "Drell-Yan"

when you discuss J production in pp collisions. Do you
mean to include only processes with an intermediate
photon? If that is the case, I think it has been known
since the time of your original experiment that Drell-Yan
cannot be responsible for J production because of the
ratio (I{pp~ e+e-X) on resonance / r;-(pp~e+e- X )
off resonance.
CHEN:
It is a question of terminology. For J production

by the Drell-Yan process, I mean to include all processes
ion i. e. both direct ct"

concerning a pair of ordinary quark-antiquark annihilat-
to J cO\lpling and one photon
intermediate states. For the continuum lepton pairs,
Drell-Yan refers to only the one photon process.
JUNG:
We actually did not use the term "Drell-Yan" for

these annihilation diagrams in our paper.
DISCUSSION 467
WHITE:
In your Brookhaven experiment did you measure the

target dependence of J or high mass dihadron production?
CHEN:
We have some data on J production with different

targets such as Be and Zn , but the A dependence has not
been looked into. However I can show you some data from
the Princeton-Chicago group.
JENNI:
The jl-beam dump experiment did not measure the

Drell-Yan process because w."'/s is too small and all of
the high mass events ( ~ 4 GeV/~~) are probably from
the tail of the J/t . Also, what mass range will your
ISR experiment cover, and with what mass resolution?
CHEN:
Scaling Fermilab results up to the highest ISR energy

shows that we can expect results onto masses as high as
24 Gev 1,,2. , with a resolution AM = 5%.
M
JUNG:
Can you comment on the ~p~> behaviour? Measuring

~P.L> of the lepton pairs gives us information about PJ..
of the interacting quarks (according e.g. to a quark-
antiquark annihilation picture or a quark antiquark fusion
picture) . In a naive parton model the ~ls of the partons
are expected to be very small. The measured <PL) is, in
comparison, very high. It rises up to about 1 GeV/~ at
a mass of about 4 GeV/c~ and then (according to the new
results from the Lederman experiment) stays constant up
to very high masses (about 12 GeV/G~). Can you explain
why <P.J.) could be so incredibly high? Is there an ex-
planation for why <~) becomes flat at high masses?
CHEN:
I haven't seen those data yet and so I cannot comment.

However, since the P~ acceptance in that experiment is
468 DISCUSSION
rather small, maybe one should wait a few months for

the results from a large p~ acceptance experiment.
MIETTINEN:
May I comment on this question of the mean trans-
verse momentum of dilepton pairs vs their mass. Figurei
shows new preliminary Fermilab data presented by Leon
Lederman at the Budapest Conference, two weeks ago. As
you can see, <ak is increasing with increasing ')'K.~..k
but it levels otf at a value of <&:>~ 1 GeV Ie. above "about
""~I' '" 4 Ge VIe. 2.
This is a very important new result. If we think
that the pairs are produced in an incoherent interaction
among the constituents of the colliding hadrons (such as
in the Drell-Yan process) then the transverse momentum
distribution of the pair reflects the transverse moment-
um distribution of the constituent in the hadrons. New
data on large p~ hadron-hadron collisions from the CERN
ISR also provides evidence that the <PJ.> of the partons
is much larger than the typical transverse momentum of
secondary hadrons, which is about 350 MeV/~.
MINGUZZI-RANZI:
I do not understand why J production via an inter-
mediate virtual photon after ~1 annihilation does not
contribute substantially to the cross section.
CHEN:
The fact that the ratio of the J signal to the di-
lepton continuum is much bigger in hadronic production
than that in e+e- annihilation shows that the direct
coupling of,; to the J dominates over the single photon
intermediate state.
JENNI:
I believe that a more sensitive way to look for
charm production in hadron interactions is looking for
semileptonic decays by looking for ~~ ':f= coincidences
from cC: ; the branching ratio l> ~ K+ l y <'OJ 20% while
the branching ratio for D -+ Klr is only ,.., 2%. This
DISCUSSION 469
1.5 -
+
>CII
1.0 f-
t t+++ 1
T
f
(.!)
::1.
::1.
'"
~
**
x
........
Pilcher et al VS = 16
0.5 -
This experiment V5 = 27
I I I I I I
0 2 4 6 8 10 12 14
m l1lL (GeV)
Fig. 1. Lederman et al: preliminary data.

470 DISCUSSION
will be done in the current R 702 ISR experiment (CERN-

Saclay-ETH Zurich collaboration).
SMITH:
On that point, Fitch sees a 2.5cr bump in the Klr

mass spectrum at Brookhaven. He looks at K1r states
produced by anti-protons near threshold for bS prod-
uction.
He should have 5 times as much data in the fall.
CHEN:
In the data I have shown but never mentioned, there

is also a 2.51> effect in the k'~'R'- and )("irt invariant
mass plots at a mass of 1.85 GeV. The Purdue-Michigan-
Fermilab group also had a 2 Q effect at that mass last
year. This effect went away, in their case, after more
running. One really can't take a 2. 5 ~ effect too
seriously.
REVIEW OF LEPTON PRODUCTION IN HADRON-HADRON COLLISIONS
A.J.S. Smith
Physics Department, Princeton University
Princeton, N.J. 08540, USA
1. INTRODUCTION
The production of leptons, especially at values of transverse

-1
momentum well above 1 fermi ,is a sensitive probe of hadron
structure at small distances. In these lectures I hope to sum-
marize in an informal manner the present knowledge of hadronically-
produced leptons, with emphasis on the general features of the da-
ta, rather than on details of individual experiments. The reader
1-3)
is referred to the many recent reviews of the subject (some
still in press at this time) for other points of view. I have bor-
rowed freely from them, and gladly acknowledge the authors. Also,
some aspects of the subject were discussed by Professor Min Chen
in his lecture at this School, and are not repeated here.
After a brief introduction and physics motivation for the

study of leptons, I shall survey in turn the inclusive processes
involving one, two and even more than two leptons in the final
state. For single leptons, which occur over a large kinematic
-q
region with a yield roughly 10 of the TI yield, some of the
puzzles of the last few years have been resolved. However, there
are still some kinematic regions in which the data themselves and
their interpretation are controversial and possibly in conflict.
471
472 A. J. S. SM ITH
In the last year much progress has been made i'n studying lep-
ton pairs. Not only is there an exciting new structure at 9.5
GeV pair mass 4 ), but also much better data on continuum pairs.
Here one observes scaling behaviour, as predicted 5 ) by Drell and
Yan, and also a striking similarity in the properties of the ~n
elusive differential cross-sections of the J/~ resonance and con-

tinuum ~-pairs of neighbouring masses. On the other hand, all
searches for non-trivial multi-muon events have been negative. In
particular, rather stringent limits have been placed upon the pro-
duction of prompt muons in association with the J/~ particle, con-
trary to what might be expected in simple pictures of charmed par-
ticle production in strong interactions.
+
Searches for leptons (in what follows the notation - will
+ +
denote e- or ~-) in collisions of strongly interaction particles
have been of high interest to high-energy physicists for many
years. This is, of course, because such leptons indicate the pre-
sence of weak or electromagnetic processes, among which are the
decays of particles stable under the strong interaction, and the
production of massive virtual photons. The discoveries of the J/~
particle and the T structure at 9.5 GeV are sterling examples of

exciting searches for final state leptons. In the future, one
hopes to see leptonic signals of charm production by decays such
as DO + vx; of heavy leptons via h + vhv, etc.; and of the
long-sought intermediate vector bosons which mediate the weak in-
teraction. These bosons would signal their presence by decays
+
w- + -V and ZO + +-.
+
Experimentally it is attractive to use leptonic signatures

because detectors can be designed which are insensitive to the co-
piously produced hadronic reaction products. There are also ha-
dronic signals of the phenomena discussed above, but severe experi-
mental difficulties arise in finding interesting states in the face
of large cross-sections for uninteresting hadronic events, and of
LEPTON PRODUCTION IN HADRON-HADRON COLLISIONS 473
the high hadron multiplicities even in the uninteresting events

themselves. For example, the J/~ might well be produced in a reac-
tion
f. P - 'J"I'" -t s~s + )(
L,.. "lti'1t+n.-7t-nO
I invite you to design an experiment that would find the J/~ in

this process, even now that we know the J/~ exists. Even better,
contemplate a search for yet-undiscovered such states (the inclu-
-31 2
sive J/~cross-section is only ~ 10 cm, as compared to the to-
.
ta 1 p-p cross-sect10n 0 f > 4 x 10- 26 cm 2 ) .
Most of the interesting new physics involves leptons of large

transverse momenta PT.
a) To probe hadron structure one needs PT 1 f- 1 (or 200 MeV/c).

b) To search for decays of a heavy particle of mass M, typical
values are PT :S M/2. For a two-body decay, isotropic in the
particle's rest frame, one has the simple distribution
aN
dc.o~e~ ... "2
"
where e* is defined in Fig. la. In terms of the transverse
momentum PT ~ ~M . e*
S1n (neglect masses of final-state par-
ticles), we get
This PT-distribution exhibits a sharp peak at M/2, as shown

by the solid line in Fig. lb. In the laboratory system, this
peak is broadened by the transverse-momentum distribution of
the parent, as shown by the dotted line in Fig. lb. The im-
portant fact to note is that a significant peaking at PT = M/2
is still observable if the average transverse momentum <PT>
of the parent is much less than M/2. Fortunately, that seems
474 A. J. S. SMITH
(a)
L
~
..
Beam Direction
/
..
/
II
Missing Momentum
In Center of Mass
Of Particle M
I
(b) I
J
3 J
I
J
I-
.'.
I
0- I
..
~ 2 I'
Z
"0 ~
,
....... , ".- .- ' I '
...' .".'
J '.
I
I
I
" ...... I
J ''
...... J
........ -....
'
...
M/4 M/2
Fig. I Transverse-momentum distributions of single leptons from

isotropic decays of an unstable particle: a) definition
of centre-of-mass angle e*; b) distribution of PT in a
two-body decay of a parent at rest (solid line); and of
a parent having its own PT distribution (dotted line);
PT distribution from a three-body or multi-body decay
(dashed line).
to be the case as M gets la~ge, and hence everyone hopes to

discover the intermediate boson by measuring large PT single
+ +
muons, looking for the two-body decays W- + ~-v.
Although di-leptonic decays of neutral particles MO + +-

can also be found by peaks in inclusive single-lepton spectra, it is
much more powerful to detect both leptons and reconstruct the in-
variant mass of the dilepton from the expression M2 (p + P )2.
+ -
For multibody "semi-leptonic" decays such as M + v + hadron(s),
one gets single-lepton transverse-momentum spectra like the dashed
line in Fig. lb. No sharp peak ~s .
seen, and unfortunately, p *
max
can be much less than M/2 if heavy hadronic combinations accompany
the v pair. For example, the charmed particle decay nO + K*~V
has p * ~ 800 MeV/c, giving transverse momenta to the leptons only
max
comparable to the typical hadronic <PT> of ~ 300 MeV/c.
Electrons and muons present quite different problems to the

experimenter wishing to detect them in the presence of large ha-
dronic backgrounds, and to select the interesting "prompt" or
"direct" leptons from leptonic backgrounds from decays of ordinary
hadrons. Let us first discuss muon detection. The principal back-
grounds are the decays TI + ~v, K + ~v. Hence one exploits the
muon's unique ability (among charged particles) to traverse large
amounts of matter without interacting. By placing many absorption
lengths L b of material, preferably of very high density, near the
a s
production target, one can absorb most of the ordinary hadrons be-
fore they have a chance to decay. Because the decay probability
per unit path length decreases as lip, and the collision length ~s
approximately constant, the fraction of hadrons decaying before

they are absorbed also decreases as lip. A lower limit for decay
background is obtained when the interacting beam is immediately
absorbed in Uranium (absorption length L b ~ 12 cm). This lowest
a s
decay probability for pions is 12 cm/(ycT) ~ 0.002/p, with p in
GeV/c. It is obvious that a fixed-target high-energy accelerator
is best for detecting muons.
476 A. J. S. SMITH
Even in the most favourable cases the background muons are

not negligible. To subtract them from the raw data to obtain the
direct muon yield, one varies the length of the decay path and
extrapolates to zero. One technique is to vary the density of
the absorber and extrapolate to infinite density. Such a plot
might look like Fig. 2. The variable density method has a rela-
tively small effect on the acceptance of the detector. Another
method would be to vary the distance between the production tar-
get and absorber, correcting for any acceptance changes induced.
An unavoidable problem presents itself -- hadron showers in the
absorber also lead to decay muons, especially in the forward di-
rection, where the full beam has to go through the absorber. At
large c.m. angles and large PT this problem is not too serious,
because the hadron yield falls rapidly with PT' and on the average
a ~ from the decay TI + ~v has only half the transverse momentum of
its parent.
The backgrounds in detecting direct electrons are quite dif~
ferent. By far the most serious is the conversion of photons into

+ -
e e pairs, where one electron is not observed. Even for an ideal
+ -
detector with no material, the Dalitz processes TI o + e e y and
nO + -
+ e e yare more than 100 times as prevalent as interesting di-
rect electron events. As in the muon case, one measures the back-
ground by varying the amount of material in the spectrometer and
extrapolating to a suitable length of "negative material" to cor-
rect for the Dalitz processes as well. Further to suppress con-
tamination of pairs, the experimenters carefully plan to veto all
events with more than one electron. The difficulties are first,
that the detector acceptance can vary when extra material is added.
Second, because Dalitz pairs so dominate the real signal (especial-
ly at low PT)' the properties of the veto counters must be very
well understood. Finally, the detectors must be able to handle
all the hadrons produced along with the electrons.
Observed
Muon
Yield
Direct
Yield
1119 (Uranium) up
Fig. 2 Method of ex tracting direct muon rates from measurements
with absorbers of various densities .
o 10 20
I I I
Tor9" scolt 1ft mel"e
Fig. 3 Apparatus of Chicago-Princeton I group.

478 A. J. S. SM ITH
Even after correction has been made for the above backgrounds,
there remain real prompt, direct leptons from dull sources; for
+ -
example, the decays of Po, wo, <Po -+ Q, Q, In v~ew of the advanced
+ -
age of the J/1jJ particle, the decay J/1jJ -+ Q, Q, must now also be
classed as a pollutant. Measuring lepton pairs yields the cross-
sections for vector-meson production -- interesting physics in it-
self. However, the copious production of prompt lepton pairs is a
limiting factor in searches for new phenomena with single leptons.
6 7)
The first definitive experiments came in 1974 from FNAL' ,
8) 9) 0
the ISR ,and Serpukhov . All groups looked near 90 in the
centre-of-mass, or x F ~ 0 (the so-called Feynman x is defined as
x F = 2 p~//S, where P~ is the c.m. longitudinal momentum, and IS ~s
the available c.m. energy). Here the background rates are lowest,
and most new phenomena should occur. Experimentally it is often
easier to measure the relative yield Q,/n rather than the more fun-
damental quantity E d 3 0/d 3 p. For high PT the Q,/n ratio is also a
natural scale factor, in that both Q, and n come from hard scatters.
Also, this ratio facilitates comparison of experiments having dif-
ferent target nuclei. The basic experimental result is simply:
-4
Q,/n ~ 10 for both e and ~ for PT ~ 1.3 GeV/c. In the forward
direction, Q,/n is not as useful a quantity as E d 3 0/d 3 p because
large numbers of pions are produced peripherally in this region;
also the lack of comprehensive n cross-section measurements makes
it difficult to extract E d 3 0/d 3 p from Q,/n.
A typical experiment is the original one of the Chicago-

Princeton (CPI) group6), whose spectrometer is shown in Fig. 3.
Basically, it consists of a quadrupole-dipole beam transport sys-
tem, interspersed with scintillator-hodoscopes and long threshold
gas Cerenkov counters, the latter to tag the various particle
species. The spectrometer has a small angular acceptance centred
at 77 mr in the laboratory (90 0 in c.m. for an incident energy of
200 GeV). They measured the muon yield for various secondary
momenta by inserting a hadron absorber between the target and the
first quadrupoles. Typical results are shown ~n Fig. 4, where
the invariant cross-section E dcr/d 3 p is shown as a function of PT'
One sees that over seven decades E dcr/d 3 p for ~+ production is al-
_4 +
most exactly 10 times the n cross-section in the same region of
phase space.
Similar results were found in other experiments. Some sum-

maries are shown in the next few figures 10 ). Figure 5 shows
the /n ratio as a function of Pr at Fermilab energies, for x F ~
~ O. A gradual increase is seen as PT increases, both for elec-

trons and muons. At the ISR, for IS ~ 52.7 GeV, the results of
two
exper~ments
11,12)
are shown
~n F~g.
6 ,where e /n .~s plotte d
versus PT' For PT > 1 GeV/c one sees similar behaviour to the
Fermilab data. For lower PT' on the other hand, ~/n increases
dramatically, according to the ACHMRN group. One should note
o 0
that their measurements are at 30 c.m. rather than 90 , where the
CCRS and Fermilab measurements were made. As we shall see, such a
large increase at low PT is difficult to explain phenomenologically.
The dependence of ~/n upon the c.m. energy is is shown in Fig. 7
for PT > 1.3 GeV/c, and in Fig. 8 for lower PT: 0.4 ~ PT ~ 0.6
GeV/c. The data of Fig. 7 are reasonably consistent and seem to
indicate a threshold at is ~ 6 GeV. The situation for low PT
(Fig. 8) is not at all clear, however, and more experiments seem
to be needed to clarify things. I point out that the low x F ' low
PT region is certainly the most difficult for experiments: the
Dalitz-decays are copious, and muon experiments are almost out of
the question. Hence we shall devote most of our attention in in-
terpreting the prompt lepton data for PT > 1 GeV/c, where at least
the experiments are reasonably in agreement.
480 A. J. S. SM ITH
\
1\ 300 GeV
p+nucleon-,u+
"\
~ 10- 34
0
"-
> Q)
(!)
I
I
\
N
E
u
10- 36
,..,: .
\
'0
"-
b
'0
W
(p+ nucleon-7T+) x 10- 4 s\

I
1,
10- 39
0 2 3 4 5 6
Transverse Momentum ( GeV/c)
Fig. 4 Differential cross-section for inclusive ~ production

as a function of PT, measured by the CP I group at
Fermi lab .
['8
3.0
.IS 1'" epl 8*~90

!
2.5 (GeV) 24. j-L- CP
24 0 e CF
!
2.0
<;t
0
x
1.5
--
t:::
~ 1.0
0.5
0
1.0 4.0 5.0 6.0
Transverse Momentum (GeV/c)
Fig. 5 The ratio tin is plotted as a function of PT' for

data at Fermilab energy. The measurements were
made at x F = O.
e:t, 'IS = 52.7 GeV

6
oACHMRN at 30
v .Lower limit from
0
4 preliminary data
I::: (new set-up)
.......
Q) .CCRS at 90
P.L (GeV/c)
Fig. 6 The electron to p~on ratio at ISR energies, vs PT
Note that the points of Ref. 11 were measured at
8* = 300, whereas those of Ref. 12 and Fig. 5 were
measured at 8* = 90.
482 A. J. S. SM ITH
o et (CCRS)
1.3< P1 < 3.0 GeV/c
,.,.-(CP)
60 0 < 8*< 100 0
e ,.,.- (CP), et (CF)
2.0 v ,.,.~ (CHPW)
,.,.+ (Serpukhov)
e+ (P-SB)
1.0
Fig. 7 Dependence of 9-/rr upon IS', for measurements with

PT > 1.3 GeV/c.
2.0r- -
v
o
x
1.0 -
Fig. 8 Dependence of 9-/rr upon /S, for measurements with 0.4 < PT
< 0.6 GeV/c.
2. THE SOURCES OF SINGLE PROMPT LEPTONS
Soon after the initial discoveries, people estimated the con-

tribution to the single-lepton yield of the "old" vector mesons
+ -
p, ~, w, which all have the rare decay modes Va + , with branch-
ing ratios $ 10- 4 The conclusion was that these decays could only
account for 20% of the yield for transverse momenta above 1 GeV/c.
The J/~ contributed substantially to the yield for 1.5 < PT < 2.5
GeV/c, but even its inclusion cannot explain more than half of the
observed leptons in this region. Two years ago here, Professor
J. Cronin calculated 13 ) the /n ratio using all data available at
that time. He had to use data from many different experiments, and
assume a form for the dependence on PT and x F of the parent vector
meson cross-sections, since they hadn't been measured in all rele-
vant kinematic regions. The method was to use the vector meson
+ -
cross-sections, let the decays Va + occur, and then integrate
over one (unobserved) lepton to obtain the inclusive single lepton
yield. The results of this integration are shown in Fig. 9. The
conclusion of Cronin was that more than half of the prompt leptons
at x F ~
were unaccounted for, even at PT ~ 1.5 GeV/c, where the
J/~ makes its maximal contribution. Similar observations were
made by Bourquin and Gaillard 14 ) , who also proposed that the sur-
feit arose from semileptonic decays of charmed particles, e.g.,
D + Kv. Their results are shown in Fig. 10, where one sees the
shape of leptons, from D, D decays needed to account for the ob-
served total lepton yield, as well as the contributions of the vec-
tor meson decays. It should be noted (Fig. 11) that DD decays can-
not explain the large increase at low PT observed at the ISR 11 ) .
Other possible sources could be the leptonic decays of a heavy

lepton, e.g., h + ~vv, or the decays of massive virtual photons in-
to lepton pairs. Such virtual photons could be produced by the
Drell-Yan process of parton-anti parton annihilation, sketched in
Fig. 12. Another source of virtual photons could be parton brems-
strahlung (Fig. l2b) as proposed by Farrar and Frautschi 1S ).
484 A. J. S. SM ITH
~
(!)
......
N
E
u
..,a.
~
b
"0
W
-39
40 O'--'--2'--3'--4L-J5L-J6L-J
Transverse Momentum
( GeV/c)
Fig. 9 Estimate by J.W. Cronin (Ref. 13) of the contribution

of lepton pairs of all known sources to the yield of
single leptons. The ratio /n is plotted versus
p for e* = 90 0 (x = 0).
T
10- 3
o Columbia -FNAL
Chicago - Princeton
Ft~~j#1
1(j4
10- 5
I::
"- 10- 4
cy Total
---
10- 5
PT (GeV/c)
Fig. 10 Estimates of Bourquin and Gaillard (Ref. 14) of the

single lepton yield due to known sources of lepton
pairs, and their phenomenological estimate of leptons
from decays of DD mesons. The estimates are compared
with results of Refs. 6 and 7.
486 A. J. S. SM ITH
~ CCRS
?9 0 CHORMN -30
PT (GeV/c)
Fig. 11 Bourquin and Gaillard's estimates are compared with

results of Refs. 11 and 12. It is apparent that the
low PT rise observed by Baum et al. is not accounted
for by DD decays.
gluon
q r
( b)
Fig. 12 Mechanisms for production of massive lepton pairs

via virtual photons. a) parton-antiparton (Drell-
Yan process); b) inner bremsstrahlung.
Finally, there could be some very dull sources not yet included
in the estimates of Cronin or Bourquin-Gaillard, such as the
o +- 0 +-0
Dalitz decays n + ~ ~ y or w + ~ ~ ~ , which have not been sub-
tracted from the observed muon yields.
As it seems there are several possible sources, how can one

gain further insight as to which one is responsible for the excess
of prompt leptons? First, one can measure the polarization of the
leptons (experimentally this is possible only for the ~ case). If
the ~'s were produced weakly, the expected parity violation would
produce a longitudinal polarization. No polarization would sug-
gest an electromagnetic source. To date there are two somewhat
16). .
conflicting results. A Yale-BNL group work1ng at Ferm1lab
(IS = 27 GeV) has measured the polarization of ~'s produced both
in the forward direction and at PT = 2.15 GeV/c. In neither case
did they find significant polarization, their results being
p 0.00 0.10 (forward direction)

and
p = 0.135 0.20 (p
T
= 2.15 GeV/c).
A Serpukhov groupI7), on the other hand, with IS = 12 GeV, has

claimed a polarization for ~'s with 2 < PT < 2.8 GeV/c. They
find P = 0.85 0.37, where the sign is opposite to that expected
in a normal V-A decay. However, their result is barely signifi-
cant statistically, and I conclude that there is no conclusive
evidence for polarization.
A second question one could ask is "Does a charge asymmetry

exist in the yields of ~+ and ~-"? An affirmative answer favours
a weak source, a negative answer suggests electromagnetic. All
data except some from Serpukhov give no charge asymmetry. The
18) + -
Serpukhov group finds ~ /~ = 1.20 0.10, again a barely signi-
ficant result -- once more there is no compelling evidence for a
weak source of prompt muons.
488 A. J. S. SMITH
A most important question is whether or not the prompt leptons

are produced in pairs. There may, of course, be different answers
for various regions of x and PT' If the answer is yes, immediately
we wish to know more about the source of pairs themselves. At
least four groups have data. Proceeding more or less historically,
12)
we have first the result of the CCRS group ,who measured the
contribution of low-mass e+e- pairs only, to the yield of single
electrons having PT > 1.3 GeV/c. They give the following 95% con-
fidence limits, as a function of pair mass, of the fraction of e/n
due to pairs.
Mass (GeV) Fraction of e/ n from pairs
0.4 < 0.064

0.5 < 0.104
0.6 < 0.178
0.7 < 0.512
+-M
0.8 P < 1.62
0.9 < 2.73
1.0 < 4.45
Their significant statement is: pairs of mass M < Mp do not ac-

ee
count for a substantial fraction of the single electrons observed
with PT > 1.3 GeV/c.
19)
The second experiment was performed by the BNL-Yale group
~ + -
at vs = 27 GeV, who measured yields of ~ and ~ mesons. They
triggered on a single ~, and then asked if a second one was pre-
sent in the event. They find, for x F > 0.2 that ~ 3.5% of their
single-~ triggers were accompanied by a sec0nd~. After correct-
ing for acceptance, they get
Accompanied
. ::: 1.1
All smgle j.1
In making the large correction for their acceptance, they have as-
sumed that the j.1-pairs decay isotropically in the dimuon centre-of-
mass frame. They also have performed elaborate calculations to cor-

rect for secondary ~ production during the hadron cascade. They
conclude: the bulk of single in the forward direction (x F >
~'s
> 0.2, PT < 0.6 GeV/c) comes from pairs. The average mass of the
pairs is <M> ~ 900 MeV, and the ~-pair yield from p, , w is
times too small to account for their signal. The resolution of
this experiment lS too coarse to see the individual vector-meson
contributions.
It is clear that an experiment was needed which had an accep-

tance much larger than the Yale-BNL group, and with resolution
good enough to measure the resonant (vector-meson) and non-resonant
contributions to the yield of lepton pairs. This would make pos-
sible a precise calculation a la Cronin of the contribution of all
pairs to the inclusive single-lepton yield as a function of PT'
with no need for extrapolated or assumed pair cross-sections.
Such an experiment was performed by a second Chicago-

Princeton (CP II) group of Anderson et al., using the former
Chicago Cyclotron as a wide-aperture analyzing magnet. This ap-
20)
paratus is shown in Fig. 13. In my lecture at the 1976 Erice
School,I described this experiment in great detail, as it pertains
21 22)
to the single-lepton question. The work has been published '
and the conclusions are unchanged. Therefore, I shall refer the
reader to these proceedings for most of the details of this work.
In a later section of my lecture today, I shall report the final
results of the CP II group for hadron-induced lepton pairs.
To summarize, the CP II group have measured the differential

cross-section for ~-pair production by 150 GeV protons on a Beryl-
lium target, as a function of pair mass M, Feynman-x (x F = 2 PL//S)'
and transverse momentum PT' For completeness, the mass spectrum
lS shown in Fig. 14. A large non-resonant contribution is evident,
In addition to that of p, w, , and J/~. The differential cross-
sections are summarized in Table 1, where the data have been fitted
to the form
490 A. J. S. SMITH
Lead Walls
r P Hodoscope
j
Spark Chambers
Target
Iron Hadron Absorber
Chicago Cyclotron
Magnet Iron Hadron
Absorber
Fig. 13 Spectrometer of the Chicago-Princeton II group.
150 GeV
P"Se--fLfL+X CP-lI
~ XF >0.15
o
~ 10
u
::l
Z
:;;
Q)
t!>
"-
I
~
~
~
b
"U .1 ("7-fLfL y)+ (w-fLfLTT")
(0""7 = O"w =O"p)
200 600 1000

M (MeV)
Fig. 14 The cross-section do/dM for producing ~ pairs,

measured by CP-II (Ref. 21) at 150 GeV. Esti-
mates of low-mass pairs from Dalitz process are
shown (dashed line).
r
m
"o-i
Table 1 Z
:0
Cross-section Cross-section o"
/nuc1eus /nuc1eon o
Mass A b c c
Region Source "F > 0.15 "F > 0.15 "F > 0 C)
GeV nb/GeV 2 /c 3 (GeV/c)-1 nb nb nb -i
oZ
Proton Production
Z
I 0.21 - 0.45 Continuum 2.67 0.53 x 10' 4.63 0.15 6.03 :!: 0.22 1470 300 340 70 1550 620 :r:
II 0.45 - 0.65 Continuum 9.03 1.8 x 10 3 4.58 0.14 4.34 0.21 800 160 185 37 620 220

o
! :0
III 0.65 - 0.93 p-w 4.69 0.95 x 10 3 3.79 0.09 2.79 0.12 960 190 220 44 510 150 o
z
Continuum 1.83 0.40 x 10 3 372 75 86 17 200 60 I
IV 1.06 0.21 x 10 3 3.93 0.28 I :r:
0.93 - 1.13 4> 4.06 0.40 1.27 38 39 9 83 25
o
Continuum 0.55 0.11 x 10 3 66 20 15 5 43 13 :0
V 1.13 - 2.0 Continuum 250 100 3.41 0.85 3.78 0.80 39 16 6.0 2.5 16 7
o
Z
C)
VI 2.7 - 3.5 J 36 12 2.08 0.26 2.94 0.32 14 5 1.6 0.6 3.3 1.1 o
r
Pion Production r
en
I 0.21 - 0.45 Continuum 2.20 0.44 x 10' 5.07 0.25 4.30 0.33 1660 330 383 76 1250 500 o
Z
II 0.45 - 0.65 Continuum 4.02 0.80 x 10 3 4.64 0.24 1.92 0.25 780 160 180 37 370 130 en
III 0.65 - 0.93 p-w 3.41 0.70 x 10 3 4.31 0.16 1.34 0.14 960 190 220 44 370 110
Continuum 1.46 0.30 x 10 3 410 82 94 20 160 50
IV 0.93 - 1.13 4> 4.90 1.5 x 10 2 3.61 0.40 1.73 0.44 160 48 37 11 70 21
Continuum 1.47 0.44 x 10 2 48 14 11 3 21 6.4
V 1.13 - 2.0 Continuum 89 53 3.2 2.4 1.33 1.0 40 4 6.2 3.7 10 6
VI 2.7 - 3.5 J 81 27 2.57 0.36 1. 72 0.38 35 12 3.9 1.3 6.5 2.2
~
'()
492 A. J. S. SM ITH
Using this fitted form, we then summed over all pairs to obtain
the inclusive single-w yield. The results, as shown in Fig. 15,
are: pairs from vector mesons alone cannot account for the ob-
served win ratio of ~ 10- 4 , consistent with Cronin's calculation;
the inclusion of the previously undiscovered low-mass continuum
changes the picture - the single-w yield at x F = 0 can be explained,
for 1.0 < PT < 2.5 GeV/c, by the measured spectrum of w-pairs. A
similar conclusion applies in the forward direction, as seen in
Fig. 16.
Since last year, many checks have been made on the data, to
be sure that the non-resonant continuum is real, and not due to
secondary W production or uncorrelated background tracks. Measure-
ments made with different target thicknesses show identical mass
spectra, indicating no secondary sources. Also, tracks downstream
of the hadron absorber from one event were matched with upstream
tracks of a different event, to estimate the background from un-
correlated tracks. Finally, careful Monte Carlo estimates were
made of background from the Bethe-Heitler production of pairs by
secondary photons. Less than 3% background from any source was
found, so we are sure this low-mass continuum is established. It
cannot be explained by conventional sources such as nO ~ WWy or
WO ~ nOww, which fail by an order of magnitude to account for it.
The solid line in Fig. 14 shows the shape of the Dalitz spectrum --
it is clear something else is needed!
. 23) .
A recent exper1ment uS1ng the SLAC streamer chamber has 1n
fact confirmed the existence of this low-mass continuum. They have
+
looked for inclusive w-pair production in 16 GeV/c n-p collisions,
and have measured the spectrum (preliminary) shown 1n Fig. 17. A
very similar shape to that measured by CP II 1S seen. The authors
also conclude that an unconventional source must be responsible.
Vector Mesons (0 ) (b)

+
Continuum
<J)
:::>
Q)
3.0
U
2.5
:::l
e - Vector Mesons
~ ""~ plus Continuum
>Q)
<.!) ,x 2.0 ---- Vector Mesons only
4tJ !
l-
----
..Cl
e
~1.5
a.
'"-a
t
~
1.0
----b
'"-a
W 0.5
~!~ ____ l_~
0.1 0
2 3 0 I 2 3
PT (GeVlc) PT (GeV/c)
Fig. 15 Contribution of ~ pairs to the yield of single ~'s,

for single ~'s at 90 0 in the centre-of-mass.
PT<600 MeVic PT<600 MeV/c (b)

1.8
Vector Mesons
1.6
,,
1
1.4
\
\
,
,, ""0 1.2
\
\
,,
.l
c
f
-u.10 3 , ,x 1.0
><
-a "- I-
"-
----b Vector Mesons>'" ~ 0.8
~
-a
Only \
,
\
0.6
\
\
\
, 0.4
\
\ 0.2
\
---- - -- ---- -- - ------
10
0.1 0.2 0.3 0.4 0.5 0.6 0.1 0.2 0.3 0.4 0.5 0.6
XF XF
Fig. 16 Contributions of V pairs to the yield of single
~'sin the forward direction.
494 A. J. S. SM ITH
1T + P-fL+fL- + anything
16 GeV/c. XF > 0.3
>Q)
::0:
o 10
~
-en
c:
Q)
>
w
(."_y,.,.+,.,.-)+(w_.,,.o,.,.+,.,.-)
CT." = 1/2 CTW
200 400 600 800 1000

M,.,.jl(MeV/c 2 )
Fig. 17 Spectrum of low-mass ~ pairs measured in 16 GeV

n-p collisions in the SLAC streamer chamber.
EI> ~ ~ ~ Ef) Ef) ~ Ef) - ~ - -ElEf)9---

.rProton Beam
L~ ----------------------------T------------------
Mo 17m.rod.
gnetic Spect
rometer
loml"-__________
100m,
Fig. 18 Multihole spectrometer of the CP I group at FNAL. The holes
are filled with large liquid scintillators. The authors
trigger on a high-PI particle in the precision spectrometer,
and obtain a rough measurement of the energy and angle of any
muon penetrating the earth on the opposite side of the beam.
The last experiment I shall discuss on the single-lepton

question, done by the CP I group2 4) at Fermilab, and reported by
Prof. Cronin at the 1976 Erice School, uses their "multihole"
spectrometer shown in Fig. 18 in conjunction with the precision
spectrometer mentioned earlier (Fig. 3). A beam of 400 GeV pro-
tons was incident on their target. Triggering on a muon of PT >
> 3 GeV/c in the precision spectrometer, they measured if a muon
also appeared in one of the large liquid scintillators embedded in
the earth, on the opposite side of the primary beam from the pre-
cision spectrometer. The authors correct for the incomplete ac-
ceptance of the apparatus and conclude, for xF = 0, 3 GeV/c < PT <
< 6 GeV/c (the PT region just beyond that measured by CP II) that
at least 30% of the single ~IS comes from pairs. This is an ex-
tremely conservative lower limit. The data are consistent with all
single ~IS arising from the ~-pairs.
At this time I shall summarize my conclusions as to the single

lepton question. It is my guess that this may be the last review
of this topic at Erice.
1) West of the Iron Curtain, for beam energies above that of the
Brookhaven AGS (IS ~ 8 GeV) , and for PT ~ 1 GeV/c, single lep-
tons come predominantly from pairs of electromagnetic origin.
The same conclusion holds for the forward direction.
2) A continuum pair contribution has been discovered, and is

needed to explain the single-lepton yields. At high PT' it is
likely to be the Drell-Yan process -- for low PT' some mecha-
nism for producing low-mass virtual photons is needed. The
Dalitz decays nO + ~~y or w + ~~TIo cannot account for it.
3) The situation in the low xF ' low PT region is unclear. There

is no strong evidence for leptonic decays of charmed particles.
No clear explanation of the ACHMNR results has been forthcom-
ing, and more experimental work would be necessary to clarify
496 A. J. S. SM ITH
the situation. The yields, and ultimate sources, of ~ and e

could be quite different here!
4) The dependence on IS is quite contradictory at IS ~ 6 GeV.
5) It will be very difficult with present accelerators to gain

further insight from single lepton experiments, although very
difficult experiments might find evidence for charm or heavy
leptons on the x F ~ 0, PT < 1 GeV/c region.
4. SURVEY OF HADRONICALLY-PRODUCED LEPTON PAIRS
In a previous lecture we had seen that it is very likely that

prompt leptons are produced in pairs of electromagnetic origin.
It is then the next order of business to survey the measurements of
these lepton pairs. As we have seen, both resonant and non-resonant
contributions are present. The comparison of the kinematic depen-
dences of these two classes should give us information on the pro-
cesses at work.
A wealth of data is now available, much of it only in the last

year. The range is enormous -- pair masses from H ~ 5 MeV to M ~
~ 16 GeV; centre-of-mass energies from a few hundred MeV to ~ 60
GeV; and transverse momenta out to more than 3 GeV/c. Also, many
+ + +
different particle beams have been used (p-, n-, K-), interacting
with targets from hydrogen to lead. Naturally there remain large
unexplored regions in this multivariable space, but many striking
trends are apparent, as we shall now discuss.
We are most concerned with these questions: How does E d 3a/d 3p

depend on mass, beam particle type, longitudinal and transverse mo-
mentum, target nucleus? What similarities, or lack thereof, exist
between resonant and non-resonant pair spectra? How good is scal-
ing, i.e., is M3 da/dM ~ f(M2/s) or is M3 d 2a/dy dM ~ f(M 2 /s), as
would be expected from constituent models?
The general shape of the pair mass spectrum is sketched in

Fig. 19a. I shall proceed through the data from lowest to highest
25)
masses. Well -- how low can you go? In the SLAC bubble chamber ,
+ -
the spectrum of Fig. 19b was produced 1n the reaction n p + e e + X,
at an incident beam energy of 16 GeV. By far the most of the pairs
o + -
come from the Dalitz decays n + e e y, shown as the solid line in
the figure. It is rumoured that an excess is seen at the very low-
est masses, but this must await more analysis.
Next, we have the results of the SLAC streamer-chamber collab-

oration, mentioned previously, and shown in Fig. 17. These show
the existence of a continuum beginning at M = 210 MeV and falling
off with increasing mass.
As mentioned before, this result confirms that of CP II

group's measurements at 150 GeV. Note that the continuum is
present in both the pion and proton-induced events of the CP II
experiment. The CP II group has now analysed some more recent da-
ta at 225 GeV, which confirm the original discovery. Their mass
spectrum is shown in Fig. 20, where one sees the p, w, as well.
As to the source of this continuum, no definitive explanation

has been forthcoming. What we do know is that Daltiz decays cannot
explain it. Both CP II and the Streamer Chamber groups have esti-
mated this, finding less than a few per cent could come from Dalitz
processes. Basically, the continuum is too broad in mass. In
Fig. 14 one sees that even if all pairs at the lowest mass are at-
tributed to Dalitz processes, by the time one 1S out to 300-400 MeV
even, the Dalitz spectrum has become very small. We have then, an
apparent flux of virtual y's produced by a mechanism not yet under-
stood quantitatively or uniquely, with a sharply falling mass spec-
trum. If one now looks in the mass region from 1 GeV to to 4 GeV,
one sees a sharply falling continuum spectrum, with the large peak
at 3.1 GeV from the J/~. The CP II data 26 ,27) are shown in
Fig. 2la where one sees that n+, n , and p beams at 225 GeV all
498 A. J. S. SM ITH
.../e+e-
-29
w
-30
-34
JI'/I
N
E -32
u
.-c: -33
E V/-L+/-L-
"C
I
"- -34 I
b I
"C
I
-; -35 I
.2 I
I
I
-36 I
I
I
I
-37 I
I
I
I
- 38 I
I
I
-39
.4 .2
Fig. 19 a) Schematic lepton-pair spectrum. At the lowest pair

masses, e+e- pairs dominate the yield of ~+~- pairs.
en
~
Z 40 ~
W
>
W
400 450 200 250

Mee MeV/c 2
Fig. 19 b) Electron pairs measured in the SLAC bubble chamber.
The curve gives a fit to pairs from Dalitz decays.
500 A. J. S. SM ITH
CPII
225 GeV Ie
p+C-f-Lf-L+X
(Preliminary)
200 500 1000

M( MeV/e 2 )
Fig. 20 Data of CP II group in 225 GeV/c p-C collisions. The

sample has high mass resolution, and confirms the
earlier discovery (Fig. 14) of the low-mass continuum.
t
This I.pl .
Hom I' 01
:;-: 0 Kluber9 el
~ ~
\fW
f
~ ~
10000
~"
j IOOOO} .00.
b)
~
u
::J
c=
.......
~ 10000
N
U
.......
>C1I
<.!)
1000
.......
..0
c=
100
10
0)
Fig. 21 Data of CP II group. a) dcr/dM for ~ pair production

in 225 GeV/c n+n-p collisions with carbon. This sample
has moderate mass resolution, but much better sensi-
tivity for M > 700 MeV than the data of Fig. 20.
502 A. J. S. SM ITH
10- 30
,,
II) ,
~
\',
.!! '. , J/'"
u
~
Z to- 31
........
<II
U
........
>III
(!)
........
N
E
10- 32 i "~Jt
u
t \' .>.... j"

II '.
0
~ 10-33 ~/'"
X
I
~ pC--/L+/L- X
'0
........ pC-/Lt/L t X
-S to- 34
Fig. 21 Data of CP II group. b) The J/~ and ~' region.

produce similar spectra. The hatched region at the bottom of the

figure indicates the contribution from various Drell-Yan calcula-
tions for p-induced ~-pairs. It is clear that the continuum below
the J/~ cannot be accounted for by Drell-Yan, but that for M > 4
GeV, the magnitude becomes consistent with the data. The J/~ and
~' region is shown 1n more detail 1n Fig. 21b for the CP II data.
The resolution of this experiment is just adequate to see the ~'.
In the mass region above the J/~ the Columbia-FNAL Stony-Brook

group4) reigns supreme. Their results for d 2 a/dM dY!y=O are shown
in Fig. 22, along with an insert of previous data in the J/~-~'
region. These results were obtained in 400 GeV p-Nucleus collisions,

+ -
p + A + ~ ~ + X; only the region near x F = 0 is seen in their
spectrometer.
The main highlight of their data, of course, is the resonant

structure at 9.5 - 10.5 GeV! A clear enhancement is observed, the
shape and width of which are inconsistent with a single resonance.
The authors have named this structure the upsilon (T). Long before
these notes are read by anyone, this structure will be much better
understood, so I don't intend to speculate on it at all. For this
lecture, the beautiful continuum spectrum, out to M > 14 GeV, is of
more direct relevance. This is by far the best measurement in the
Drell-Yan region, and probably provides the functional form of the
scaling function f(M 2 /s). In a few months this group should have
similar data with 200 and 30Q GeV proton beams, and be able to make
a definitive test of the scaling hypothesis.
Let us now try to summarize the dynamical features of lepton-

pair production. For each beam particle type we have to study the
following:
1) Alignment: in any frame we can look at the angular distribu-

tions of the decay muons. The usual simple method is to write
da/dQ * ~ 1 + a cos 2 e* and fit the data to the best value of a.
We are most interested in events with M >2 GeV, since for lower
504 A. J. S. SMITH
3000
NUMBER OF
EVENTS WITH
2.5 <M<5GeV
1000 1250A
500
300
100
50
Ii' 3
-)8
10
6 10 12 1"1 /6
}foss G.... V
Fig. 22 Results of Herb et al. (Ref. 4) for d 2 a/dH dyly=o
in 400 GeV p-Nucleus collisions. The inset shows
earlier results for lower pair masses.
masses, many production mechanisms are involved, and are certainly

not understood. Because the total variation in dcr/dQ * is only a
factor of 2, even if a = 1, a large and well-understood acceptance
is essential, as well as a huge data sample. So far, the only sig-
nificant measurement is that of CP II for the J/~ itself. The data,
shown in Fig. 23, are consistent with no alignment: a = -0.28
0.22 measured in the helicity frame. There is no "best" frame
a priori in which to do these measurements of a. Also, the PT of
the quarks could smear out the 1 + cos 2 8* alignment expected if
the process were simply qq annihilation.
What would really be desirable is a sufficiently large sample

of continuum pairs, where it is more likely that the qq annihila-
tion is responsible. Such data should be forthcoming from the next
CP II experiment, which is presently being set up at Fermilab.
2) ~F distributions. Here, the most comprehensive results come

from the 225 GeV run of the CP II group. A typical data sample is
shown in Fig. 24a for the J/~, and Fig. 24b for the M ~ 1 GeV re-
gion. One notes the general feature that the pion-induced events
have a broader distribution in x F than those induced by protons.
This feature persists from M = 200 MeV up to M = 3 GeV, a result
not easily understood. Another feature is that dcr/dx for TI and p
seem to be about equal as x F + O.
N
Quantitatively, the data have been fit to a power law (1 - x F ) .
The power N is shown as a function of pair mass M in Fig. 25. Vir-
tually no dependence on M is seen. The value of N for proton beams
remains consistently higher than N for pion beams - we see approxi-
mately that N ~ N + 2, a prediction of many models. What is sur-
p TI
prising is that this behaviour persists down to low masses, and
that it applies both to resonant and non-resonant ~-pairs.
3) RT distribution. Here the data come mainly from CP II, CFS,

and, for neutron beams, from the Columbia-Fermilab-Hawaii-Illinois
collaboration 28 ) (CFHI). The major qualitative conclusions are:
506 A. J. S. SM ITH
p-w MASS, 17":1: on e (el
--- ---------7
1.2
1.0
.8
~
.6 1+ 0.02 cos 2 (J "\
'c
:::> .4 \
IT1
....
>- .."
,
EFFICIENCY-----,\
c .2
.... \ ~
..... ()
:0
.... IT1
z
5 J/'i' MASS, 17":!:,p on e8 Sn ()
1.2 -<
<J:)
en 1.0 1.0
0
u .8
"0
....... .6
b
"0 .4 ----- "
.....
.2 "
EFFICIENCY~",
0 "
e
0 .2 .4 .6 .8 1.0
cos
Fig. 23 Distribution in polar angles of ~'s from J/$ + ~~

and pO + ~~.
2.7< m < 3.5 GeV/c 2

C TARGET
Fig. 24 Distributions da/dxF tor p + C + ~~ + X,

at 225 GeV/c (From CP II group). a) The
J/~ region (2.7 < M < 3.5 GeV)
508 A. J. S. SMITH
.95 < m < 1.1 GeV/c 2

C TARGET
(f)
~ 103
W
....J
U
~
Z
......
(f)
Z 102
fl::

CD
0
z

z
> 103
~
(!)
lL.
X
"0
......
b
"0 102
W
XF
Fig. 24 b) The M ~ 1 GeV region.
LEPTON PRODUCTION IN HADRON - HADRON COLLISIONS 509
I I I
E dO" _ (1-X )N
dX F F
pC
7 o rr+C
6 rr - C
5
N 4
3 t
f ! J
J I t
2 t ( i
~ ~
~
~

I
1 2 3
M (GeV/c 2 )
Fig. 25 The dependence of the power N as a function of pair mass,

where we have fitted the xF distributions to Edcr/dxF ~
(1 - xF)N .
510 A. J. S. SMITH
first, the PT-dependence is not a function of either the beam par-

ticle type or of target atomic number. This is in contrast to the
n-proton differences in the xF distributions. On the other hand,
the breadth of the PT distributions is mass dependent. A typical
PT distribution is that of the J/~, shown in Fig. 26. The cross-
section da/dp~ falls off roughly exponentially with PT' In order
to be less dependent on the function used to fit the data, we show
simply the average transverse momentum as a function of mass in
Fig. 27. The data for M < 4 GeV come from CP II, for M > 5 GeV
from CFS. A linear rise is seen up to M ~ 4 GeV. Thereafter
<PT> is independent of M. One can interpret this limiting <PT>
of about 1.2 GeV/c as the measurement of the <PT> of the quarks
themselves. This is another beautiful result of the CFS experi-
ment.
4) Dependence on target atomic number. There are three experi-

ments (CP II, CFHI, CP I) which have explored the atomic-number
cross-sect10n
d epen d ence 0 f t h e ~-pa1r . 29- 31) Express1ng
.
a (A) ~ Aa. ,
one would naively get a. = 2/3 for an opaque spherical target, and
a. = 1 for a transparent target. The data are shown in Fig. 28,
where the power a. is plotted versus pair mass. In the pO mass
region, we see a. ~ 2/3, as expected for an ordinary hadronic pro-
cess. The power increases with mass so that a. ~ 0.9 at the J/~.
The absorption of the J/~ itself has been taken into account (it
is very small). Within large errors, the higher-mass data of CP I
are consistent with a transparent nucleus. As I understand this,
low-mass hadrons are produced by highly-probable soft scatterings.
High-mass objects, on the other hand, come from hard scatterings.
The fact that the beam particle has previously undergone an ordi-
nary low PT interaction does not remove it from the "beam", and it
(or its fragments) can still undergo a hard scatter further into
the nucleus. Although not shown in the figure, a similar picture
has been observed by CP II for pion-induced pairs.
2.7< m < 3.5 GeV /C 2

103 C TARGET
2
"-
(/)
z
a::
~
CD
o
Z
.~
Z 10 2
Q.-i
'0
"-
b
'0
10
Q.-i
"-
w
1.0
Fig. 26 Differential cross-section da/dpf vs PT

for J/~ production at 225 GeV/c (CP I I
group).
512 A. J. S. SM ITH
1.5 o BRANSON el 01
Do HERB el 01
0

i
'"> ( ~ ~
Q)
(!)
1.0 ~
d!
A-
~
v I
0.5
2 4 6 8 1O
M (GeV/c 2 )
Fig. 27 The average transverse momentum <pr> of ~ pairs as a function

of pair mass. The data are from CP II and from Ref. 4.
I I I
1.0
0.5 -
Branson et.ai.
x Binkley et.a!.
o Hom et.a!.
DAntreasyan et.a!.
o 2 3 4 5 6 789
MfLfL (GeV/c 2 )
Fig. 28 The dependence of ~ pair cross-sections upon

atomic mass number A. The cross-sections are
fitted to the form o(A) ~ Aa, and a is plotted
vs M.
5) Dependence on beam-particle type. The effects of beam-particle

type on the xF ' PT and A dependences of pair production have al-
ready been mentioned. Another question is simply, what is the
ratio of the total cross-sections for various beams, as a function
of pair mass?
These ratios are important tests of the Drell-Yan model. The

issues were discussed in my lecture last year, and will only be
briefly outlined here. If ~ pairs are produced by annihilations
of valence quarks, then pions should be much more effective than
protons in producing them. Also, since the annihilation is propor-
tional to Q2, where Q is the charge of the annihilating quark, up-
quarks (Q 2/3) should have four times the cross-section as down-
quarks (Q 1/3). For an isoscalar target (carbon) we get the pre-
diction a(n+C + ~~ + X)/a(n-C + ~~ + X) = 1/4. Unfortunately, the
sea quarks dilute this ratio, particularly for small values of
M2/S. Numerous authors have calculated this effect 32 ,33).
The CP II data, along with the predicted ranges of the ratio,

+ -
are shown in Fig. 29, where n /n , and p/n are also shown. The
n+/n- ratio is very suggestive - it is unity at low masses, where
isospin-conserving strong interactions are undoubtedly responsible
for the ~-pairs. Similarly, it is close to unity in the J/~ region.
However, the mass bins surrounding the J/~ show the predicted asym-
metry. If this result holds up with larger statistics (CP II will
have> 20 times as much data ~n the next few months) it is perhaps
the most conclusive evidence for the Drell-Yan mechanism. One
notes that the n/p ratios are also consistent with Drell-Yan pre-
dictions. At higher values of M2/s the predicted n/p ratio becomes
dramatically large 34 ), as shown in Fig. 30. At M2/S ~ 0.2
(M ~ 10 GeV for a 225 GeV n beam) pions should be more than one
hundred times effective as protons. (The CP II group will also
measure this n-/p ratio out to M ~ 10 GeV, in its upcoming run.)
514 A. J. S. SMITH
1
p-w
-+-t -+- ++ t J/Ij!
AI
LL
X
~
'01
......
b
o
~
0::
~5~--~--J---~2~--L-~3~--L---
M (GeV/c 2 )
Fig. 29 Ratios of 225 GeV ~ pair cross-sections for

various beam particles (CP I I group).
11'- p-p.: fL- + x

PP -fL+fL- + x
100
10
o .05 .10 .15 .20 .25 .30

m 2 /S
Fig. 30 Prediction of the Drell-Yan calculation of

C. Quigg (Ref. 34) for the ratio
(n-A 7 ~~ + X)/(pA 7 ~~ + X) as a function
of M2/S.
LEPTON PRODUCTION IN HADRON-HADRON COLLISIONS SIS
6) Dependence on beam energy and evidence for scaling. At the

time of writing, it appears that the CFS group will shortly have
excellent ~-pair data for 6 < M < 14 GeV, for beam energies of
200, 300, and 400 GeV. However, today we only have their 400 GeV
data, and must compare it with other experiments to test scaling.
Clearly, continuum events are those expected to obey the law
M3 d 2o/dM dyio = f(M 2 /s). The J/~ or T do not have to scale in
35)
this way. Mel Schochet has compiled the data very recently ,
and what I now have to say comes mainly from him. A grand plot of
continuum cross-sections 36 ) is shown as a function of T M/IS =
in Fig. 31. Reasonably good agreement IS seen. The CFS data at
400 GeV give the shape of the function f(M 2 /s). It is well fit by
an exponential:
'2.
-ft~') = C e~p (-14-.7 M )
s .Ji
7) Comparison of resonant and non-resonant pairs. The J/~ data

are plotted 37 ) in a similar way in Fig. 32, where Bdo/dyio is shown
as a function of M/IS. As noticed by Schochet, all J/~ measure-
ments to date fit on a "scaling" curve having the same exponential
shape as that of the continuum. It is as if very similar mechanisms
are at work, differing only in coupling strengths. Within the
large experimental uncertainties one sees an analogous behaviour
for the ~'. We can get the dimensions of the resonance curves
(Fig. 32) the same as the continuum if we multiply the J/~ and ~'
cross-sections by M2. Then if we compare do/dy, not Bdo/dy, we

find an intriguingly simple result:
1
200
516 A. J. S. SMITH
x Ref. 36 (Upper limits)
o Ref. 34
Ref. 4
N
> Q) ~0-33
t9
I
N
E
u
>.
"C
~
"C
.........
b 10-35
N
"C
ro
~
~-~L-__~__~____~__~____~__~____~
o 0 .2 0.4 0 .6
M/.IF
Fig. 31 Survey of measurements of d 2 o/dm dyly=o,

taken from Ref. 35.
I, 0
x
Ref 3
Ref. 36
0 Ref. 26
1032
tIll ~ "
Hom et al (Ref. 34 )
Antipov et al (Ref. 37)
Nagy et al (Ref. 37)
(; CCRS (Ref. 42)
N ~0-33
E
u ,, i J/t
0
,,
"0
>.
46 34 "1,
f
If,,,
........
b
"0
ro
~O-35
,,
\ ,
~0-36 \
0 0.6
Fig. 32 Dependence on beam energy of Bda/dy for J/~

production in hadron collisions. This curve
has a very similar shape to that of Fig. 31.
518 A. J. S. SMITH
Comparing J/~ and ~' ~n this manner we get
t<'\2 dO" /d'1 ' :1/'1' ~ b

",,2 dold~ \ -\I'
a substantial suppression of ~' production, as might be predicted

by the gluon-fusion model of J/~ production, where first we have
gluons fusing to form L = I cc states, which decay via photon
emission to the J/~. Since there are no such states above the ~',
its production is suppressed.
Finally, we look at other features of resonant and non-

resonant pairs: the x F distributions, PT distributions, and A de-
pendence behave very smoothly as one passes through the J/~ mass
region, as seen ~n previous figures and discussion. This behaviour
is also consistent with the J/~ being formed in some process by
which constituents annihilate. To gain further insight, experi-
ments such as measuring what other particles accompany the J/~ are
crucial; for example, one should look for photons expected from
the decays of the L = I cc states.
5. SEARCH FOR MULTlMUON EVENTS
The narrow width of the J/~ is accounted for, empirically, by

.. 38). .
the Okubo-Zwe~g-I~zuka Rule , wh~ch states that quark l~nes have
to flow from initial to final state, and that any processes are
suppressed wherein quarks must annihilate and then reappear as a
pair with a different flavour. Since_MJ/~ < =~, the cc in the
J/~ must annihilate and change into uu, dd, ss for hadronic decays
to take place. Similar suppressions in the production of J/~'s
should be operating for OZI forbidden processes: the process
TI + A ~ J/~ + non-charmed hadrons
should be less frequent than the associated production process

n + A + J/~ + D + D + anything
One can search for the latter process by measuring the rate of
+ -
three-muon events, two muons coming from the decay J/~ + ~ ~ , the
third from the semi-leptonic decay DO + K~v, which has a 10% branch-
ing ratio. The background is, of course, muons from decays of
ordinary hadrons produced along with the J/~. As described earlier
in this lecture one tries to absorb as many hadrons as possible be-
fore they decay. The CP II group reached a limi t 39)
o(J/~ + ~ + anything) < 2.5 x 10- 3 with 90% confidence.

o(J + anything)
Folding in the D + K~v branching ratio, and a reasonable assumption

as to the model of J/~ DD production, they find
o(J + charm)
O) < 0.01 w~th 90% conf~dence .
o (J + anyt h ~ng
A similar limit has been reached by the CFHI group40), for an

incident neutron beam. The simple conclusion from all searches so
far is that no multi-muon events have been observed. This is con-
sistent with the negative results of other searches for charm in

hadron~c
coll~s~ons
0 41) h
, where teD + Kn
s~gna 1 1k
was 00 e d f or.
Acknowledgements
The author wishes to thank Professor Zichichi for the oppor-

tunity to participate in this School. As usual, it has been a
most pleasant and stimulating experience. These lectures are
based mainly on the work of the CP II collaboration, the members
of which are K.J. Anderson, J.G. Branson, G.G. Henry, G.E. Hogan,
K.T. McDonald, C.B. Newman, J.E. Pilcher, E.I. Rosenberg,
G.H. Sanders, and J.J. Thaler.
520 A.J.S.SMITH
REFERENCES
1) P.A. Piroue, Annual Meeting of the Division of Particles and

Fields, American Physical Society, Brookhaven (1976).
2) B.G. Pope, Palermo Conference, European Physical Society (1975).
3) S.C.C. Ting, Proc. of the 1976 Erice School on Subnuc1ear

Physics.
4) S.W. Herb et a1., Phys. Rev. Letters 39 (1977) 252.
5) S.D. Dre11 and T.M. Yan, Phys. Rev. Letters 25 (1970) 316.
6) J.P. Boymond et a1., Phys. Rev. Letters 33 (1974) 112.
7) J.A. Appel et a1., Phys. Rev. Letters 33 (1974) 722.
8) F.W. Busser et a1., Physics Letters 53B (1974) 212.
9) V.V. Abramov et a1., reported in Proc. of the 17th Int. Conf.

on High-Energy Physics, London, p. V53 (1974).
10) These figures are taken from Ref. 1), where direct references
to the individual data can also be found.
11) L. Baum et a1., Physics Letters 60B (1976) 485.
12) F.W. Busser et a1., Nuclear Phys. Bl13 (1976) 189.
13) J.W. Cronin, Proc. of the 1975 Erice School on Subnuc1ear

Physics (Plenum Press).
14) M. Bourquin and J.M. Gaillard, Phys. Letters 59B (1975) 191.
15) G. Farrar and S.C. Frautschi, Phys. Rev. Letters 36 (1976) 1017.
16) L.B. Leipuner et a1., Phys. Rev. Letters 36 (1976) 1011;

M.J. Lauterbach et a1., Phys. Rev. Letters 37 (1976) 1436.
17) N.Z. Anisimova et a1., Phys. Letters 65B (1976) 85.
18) V.V. Abramov et a1., Phys. Letters 64B (1976) 365.
19) L.B. Leipuner et a1., Phys. Rev. Letters 35 (1975) 1613;

H. Kasha et a1., Phys. Rev. Letters 36 (1976) 1007.
20) A.J.S. Smith, Dimuon production by pions and protons at

Fermi lab , Proc. of the 1976 Erice School on Subnuc1ear
Physics.
21) K.J. Anderson et aI., Phys. Rev. Letters 37 (1976) 799.
22) K.J. Anderson et aI., Phys. Rev. Letters 37 (1976) 803.
23) K. Bunnell et a1., SLAC preprint (Streamer Chamber Group)

(1977)
24) D. Antreasyan et a1., Phys. Rev. Letters 37 (1976) 1451.
25) J. Ba11am et a1., private communication.
26) J.G. Branson et a1., Phys. Rev. Letters 38 (1977) 1331.
28) B. Knapp et a1., Phys. Rev. Letters 34 (1976) 1044.
29) J.G. Branson, Ph.D. Thesis, Princeton University (1977).
30) M. Binkley et a1., Phys. Rev. Letters 37 (1976) 574.
31) D. Antreasyan et a1., Phys. Rev. Letters 39 (1977) 252. A

point at 6 GeV with large errors has also been measured
by D.C. Hom et a1., Phys. Rev. Letters 37 (1976) 1374.
32) G. A1tare11i et a1., Nuclear Phys. B92 (1975) 413.
33) G.R. Farrar, Nuclear Phys. B77 (1974) 426.
34) C. Quigg, private communication.
35) M.J. Schochet, review of di1epton production at high energy in

hadron-hadron collisions, Stanford Summer Institute, July
(1977). Figures 29 and 30 are ~aken from this report.
36) All data on this plot have been referenced previously except
that of J.H. Cobb et a1., Phys. Letters 72B (1977) 273.
37) Data points not previously referenced are those of E. Nagy et

a1., Phys. Rev. Letters 36 (1976) 1415;
Y.M. Antipov et a1., Phys. Letters 60B (1976) 309.
38) A. Okubo, Phys. Letters 5 (1963);

C. Zweig, CERN Report TH.412 (1964);
I. Izuka, Progr. Theor. Phys., Supp1. No. 37-38 (1966) 21.
D. Sivers, Phys. Rev. D11 (1975) 3253;
R.M. Barnett and D. Silverman, Phys. Rev. D12 (1975) 2037.
522 A. J. S. SM ITH
40) M. Binkley et a1., Phys. Rev. Letters 37 (1976) 578.
41) D. Bintinger et a1., Phys. Rev. Letters 37 (1976) 732; see

also Ref. 3).
DIS C U S S ION
CHAIRMAN: Prof. A.J.S. Smith
Scientific Secretaries: A.B. Carter, L. Rolandi
DISCUSSION No.1
PAAR:
I believe I see a contradiction among three pieces of

of information you showed this morning: First, you show-
ed that the Drell-Yan calculations agreed with the
Columbia-Fermilab dimuon mass spectrum. Second, the
Chicago-Princeton group concludes that at least 70 per-
cent of their single muons are from pairs. Third, and
here is the contradiction, you showed a plot from the
Chicago-Princeton group of the invariant differential
cross section versus transverse momentum for single muons
which was far above the Drell-Yan curve shown on the same
plot. My question is: how can the Drell-Yan prediction
for dimuons agree with the Columbia-FNAL data but not
with the Chicago-Princeton single muons if these single
muons come from pairs? This data is all from the same
transverse momentum region.
SMITH
I did not do these Drell-Yan calculations, but I be-

lieve that they do not take into account the parton trans-
verse momentum distribution. The Columbia-FNAL data was
a plot of cross section versus mass in which the trans-
verse momentum was integrated over, so this deficiency
of the Drell-Yan calculation did not show up. The
Chicago-Princeton data, however, was a plot of cross
section versus transverse momentum. We would not expect
523
524 DISCUSSION
a Drell-Yan calculation which did not take the transverse

momentum dependence of the parton distributions into
account to agree with the data on such a plot.
But I do not know the details of the Drell-Yan cal-
culation. I will try to include a discussion of this in
my lecture notes.
I might add a remark in general on the question of
comparing dilepton and single lepton transverse momentum
distributions. In order for such a comparison to be mean-
ingful, both sets of data must extend over comparable
ranges. A parent particle with 'P J.. of 3 GeV, for example,
will contribute to single leptons having ?J.. up to about
3 GeV, and the range will depend upon the parent mass.
You cannot compare the single lepton and dilepton data
unless the p~ of the daughter leptons from measured pairs
are in the same range as the single-lepton experiments.
PAAR:
To get back to the Drell-Yan calculation, I believe
that color was not included in that calculation.
ANTREASYAN:
I might be able to clear up this question. The

Drell-Yan calculations you show are old. More recent
calculations at SLAC, which include both color and the
observed parton p~ distributions, are in agreement with
the data. We at Chicago have even extrapolated these
down to the low-mass region in a naive way, and they
agree remarkably well even in this region.
MIETTINEN:
At Budapest a few weeks ago, Kajantie summarized

this question, and he confirmed that Drell-Yan is in
excellent agreement with all data.
WHITE:
You showed a graph comparing the lepton-to-pion data

as a function of Pi to a calculation of Bourquin and
DISCUSSION 525
Gaillard in which they summed up various known contribu-

tions to the spectrum. They found that if they excluded
the estimated contribution from charm, their calculated
spectrum lay significantly below the data. I understand
that at that time some people explained this discrepancy,
or gap, between prediction and experiment as the contri-
bution of eta Dalitz decay. This would have been an
acceptable explanation of the muon spectrum, but it
could not be simultaneously applied to the electron
spectrum because this source, i.e. eta decay, is a well-
known background in electron experiments and is supposed-
ly subtracted away. Therefore eta decay could not explain
the gap between the data and the Bourquin-Gaillard pre-
diction in both the muon and in the electron spectra:
there must be another contribution. Therefore, the
recent SLAC streamer-chamber data, which shows a size-
able continuum contribution at low p~, should not have
come as a surprise, should it?
SMITH:
The gap between the Bourquin-Gaillard calculation

and the data extended out well beyond 1 GeV p~, and no
one would claim that eta decay contributes that far out
in p~. In our calculation, the low-mass pair data only
contribute to the very lowest transverse momenta (i.e.,
below 500 MeV), even though it represents the bulk of the
dimuon cross section.
As to the charm contribution, a sizeable contribut-
ion is ruled out by the close match of the dilepton and
single lepton fluxes.
CHEN:
Data now indicate that D ... t<.~ ~ \'> dominates

D --'> ~ IJ\? so the contribution of these muons to dcr/df>~
is limited to very low p~.
HINCHLIFFE:
Yes, the calculation of Chris Llewellyn-Smith and

myself indicates that D decay gives a very soft lepton
spectrum but cannot explain the HPWF points.
526 DISCUSSION
SMITH:
We cannot say anything before we know the p~ distri-

bution of D's produced in hadron collisions, and this has
not been measured. In conclusion, we have no muon data
for p~ less than 1 GeV. A muon measurement is perhaps
not as difficult as an electron measurement because you
do not have troublesome background from Dalitz decays or
other virtual photon sources. However, even muon exper-
iments at low p~ are very hard because of the presence
of j(. ~ p.~ and 1(. ~ p.'I> decays.
HINCHLIFFE:
Is any muon experiment planned?
SHITH:
I do not know of any.
CHEN:
We have electron data in this region, but we have

not had time to analyze it, and such an analysis would
be difficult and model-dependent.
HIETTINEN:
I wish to mention one empirical observation which

mayor may not have significance. The It .... 1t. - mass dis-
tribution in high energy (FNAL, ISR) small- P.L hadron
collisions has the same peculiar shape as the \.l+ \J.. - mass
distribution you have just shown, i.e. there is a large
continuum which peaks very close to the threshold.
If you look at the difference of the 1(T and n .... n T
mass spectra the effect is even more striking.
JUNG:
You mentioned that you use in your experiment

(Chicago-Princeton) chambers in front of the absorbers.
These chambers are very important for your resolution.
Can you say a few words about that? To what extent could
you really use these chambers?
DISCUSSION 527
SMITH:
6
We had 10 part/sec: that leads to 3-4 10 part/sec
in the chambers. The beam spot was about 2.Sx3 cm 2 and
the electronic dead time was about 500 ns, so from this
side things turned out well (after a while). A problem
we had foreseen was possible back-scattering from the
absorbers, but it turned out that there was no serious
back-scattering. So, once we understood this, it caused
no problem. About half of our data include the inforrr~
tion from these chambers, and that information is really
only important for low-mass resolution.
WHITE:
I have a comment on your graph showing the s-depend-
ence of "low-pJ." lepton production, where "low-p",," is
supposed to mean 0.4< p",,< 0.6 GeV/c. In CCRS, our lowest
p~ bin starts at 0.6 GeV/c, so our point should not really
be included. If the lepton-to-pion ratio is increasing
below 0.6 GeV/c, it would clearly be wrong to do so.
SMITH:
I did not mean .4 to .6 GeV/c as a bin. I was only
indicating the approximate domain of PJ. from which these
data are taken.
DISCUSSION No.2 (Scientific Secretaries: L. Rolandi,

S. White)
CHEN:
The huge ~ peak in the dimuon mass spectra looks
very familiar to me. In a 1969 experiment on photoprod-
uction of e+e- in the f> ' ro region we found that the CO
peak is more than a factor of 2 greater than the r
whereas we know that ",-2: Yi2.= 9: 1. So here it's
not necessary to have m~re ~than p produced but we can
understand this as an interference effect.
You showed a scaling curve for the J and the ,.+p. con-
tinuum where the two sets of data were the same except
for a factor of 200. To me this implies that vector mes-
on (~. Col. C, :J ) production mechanism and the Drell-Yan
528 DISCUSSION
process (mediated by a photon) are very similar and could

give rise to interference effects. Isn't it possible
that just as we were able to explain the ~shape as a
~-c.o interference and the <p shape from the interference
with Bethe Heitler, we can explain the Y shape as arising
from interference with the Drell-Yan continuum.
SMITH:
I think it may be possible to do this in principle

but very hard in practice. We just don't know as much
about this process as we do about Bethe Heitler. Also,
in the photoproduction experiments, the coherence of the
process meant that an exclusive final state was measured.
ZICHICHI:
To me the most interesting interpretation is that

this structure is composed of more than one state,
like "J'/'P, \\)'
PREPARATA:
I think it is clear that it cannot be a single re-
sonance. We can estimate the expected rate for r (1'... ~tl)
and we have this measurement of 1.2 GeV for the total
width. So its branching ratio would be so small
(10-+- l~s ) that it is unlikely that CFS would have seen
it. On the other hand if it is anything like the r it
is very reasonable to expect that there will be a total
of 2-3 states with small separation.
MIETTINEN:
I think another question we must keep in mind is

whether bb and t ~ make "'-like systems which are very
near in mass.
SMITH:
This all is just speculation until we have data

from e+e- colliding beams.
DISCUSSION 529
WU:
Can you make a consistent picture of the A-depend-
ence of the dimuon production spectra and the A-depend-
ence of inclusive single muon production as measured by
CP-l?
SMITH:
CP-l single muon measurements extend to very high

?~, where the A-dependence of ~ production is larger
than A i.o Our measurements of dimuon production don't
go out this far in ~~
ANTREASYAN:
Though we published a systematic study of inclusive

hadron production and its A-dependence we have no firm
numbers on the A-dependence of single muon production.
SMITH:
So I would summarize by saying that an A-dependence

exceeding A 1.0 has only been seen in hadron production.
WU:
I would like to have a clear picture of the J prod-
uction mechanism. Can you explain the relative shapes
of d~(?f'~:n<.)/dxl=' and dCT\1t"I>~:rX)/dxF ? You
showed that at small x these cross sections are approx-
imately the same but at large x l=' the "It. cross section
becomes larger.
SMITH:
To understand this you should look at the relations:
Xb-XT=X F
2
')tbXT =.!!...
s
'::1 9/400
for our experiment, where Xb is the x F of the beam par-
ticle parton and x T is the x F of the target parton.
530 DISCUSSION
The difference between the x of the partons give

the x F for the J'. The product is 9/400, a small number.
This means that at least one of the partons must come
from the sea. So the fact that 7t carries a valence
antiquark doesn't make all that much difference (at xF
=0). It would probably make a difference at OMEGA where
s is small and valence antiquarks can play a dominant
role. I think you can naively understand the difference
at large x F because the pion has fewer partons and there-
fore they carry more momentum on the average.
JUNG:
I am still confused about the interpretation of the

p~ measurements in dilepton production. Some people
tell me the high <?~~ values are very hard to understand.
Professor Miettinen tells me that he understands those
high values and the new limitation at 1.2 GeV. Now you
come out with this Constituent Interchange Model which
explains this from the high Pol. values of interacting
quarks. So I ask the experts to clarify the situation.
SMITH:
For the naive parton model we would say the lepton

pairs are produced by fusion or annihilation of quarks
in the hadron collisions and so their transverse moment-
um should have a hadronic ~~distribution. This would
not agree with the growth in <p.l.)that we find as a
function of dilepton mass.
In the CIM model we are told to distinguish (I'm
quoting Glennys Farrar) between whether transverse mom-
entum of the dilepton system is characteristic of some-
thing that was originally there before the collision
took place or if it comes about during the hard scatter-
ing that gave rise to the formation of the pair. So that
in this picture, unlike the naive parton model, you could
probably get average <?~as high as 1 GeV. I cannot ex-
plain why this model would predict a levelling off at
about 1.2 GeV. Perhaps Professor Miettinen could explain
this.
MIETTINEN:
In hard scattering the time scale is so short that

the interaction occurs without the participation of other
DISCUSSION 531
partons. This means that transverse momentum is locally

conserved and the transverse momentum of the pair is eq-
ual to the new transverse momentum of the interacting
partons. To get <P~7 you have to convolute the parton
transverse momenta and this will give something like
<P~> "'~l:' ~ 2 <r1)pa<"\on. This is very interesting
because it means that we have measured for the first time
the transverse momentum of the partons. From the large
values of ( Pi-parton) we can say via the Uncertainty
Principle that partons are very localized in the hadron.
At low mass this process is dominated by hadronic
resonance production so this is not really relevant to
what lIve said and the small average is not surprising..
PREPARATA:
But this has nothing to do with the Constituent

Interchange Model.
MIETTINEN:
I agree. It is not necessary to go to the Constit-

uent Interchange Model to understand these features.
ZICHICHI:
Who is telling you what the mean transverse momentum

of the partons is going to be? Is there a theoretical
prediction?
PREPARATA:
I think people were surprised to find a number as

high as 700 MeV for the parton transverse momentum.
BACE:
I would like to comment about the persistent claims

that the parton p~ must be as small as 300 MeV. This is
really too naive. If the x's coming out of a hadronic
collision carry such a PL surely the parent quarks must
have had a larger ~L. Arguing this in detail Levin and
Ryskin have found a parton <Pi"> N 1 GeV a few years ago.
532 DISCUSSION
Their paper includes six reasons for why the partons

actually carry <p~) from 1 GeV to 1.S GeV.
HINCHLIFFE:
If you believe in scale invariance you would expect
that <~L~ would be directly proportional to Q2 - the
only other scale in this picture. So you would expect
that <?L> would continue to grow with mass. However, as
Feynman has shown in his book, this also can depend on
R = a L /o-T and you get the relation <?J.'2.) ~ ~z'R.
When you calculate R in an asymptoticallY free gauge
theory, R(Q2) falls rapidly as Q rises so that
< 1>l"> ac Q2 'R.(Qt) can start to level off.
PEREZ-Y-JORBA:
It seems that one can fit the cross section
M? d~/dmd::J I ~=o for the production of lepton pairs
by very different functions of M2/S J one being exponent-
ial in M'ls the other one exponential in ""/Ji. Could
you comment on this? I also would like to know whether
there are theoretical predictions on this functional
form.
SMITH:
First of all if you look at the Chicago-Princeton
data I think you will agree that the error bars are
probably large enough to accommodate a fit of either
type. The new and more precise measurements of CFS can
probably not fit the form ex\, l-a. MVS). At any rate
the agreement between the data and the form ex~ l-b M/.JS)
is very striking to me. There are probably theoretical
predictions derived from parton distributions measured
in deep inelastic lepton-hadron scattering for the
distribution, but I think at this time the fits are
purely empirical.
NARRm-l RESONANCES IN BB REACTIONS
S. Nilsson
Institute of Physics, University of Stockholm, Sweden
The experiments investigating resonances produced in

antibaryon reactions have recently made some remarkable dis-
coveries. New theoretical ideas might also be the clue to the
understanding of the phenomena. The characteristic feature of
the new resonances are that they are mesons, but with narrow
-
width and strong coupling to BB-states. The experimental situ-
ation will be reviewed in this report. The content of the re-
port will be organized as follows 1. Introduction, which in-
cludes a discussion of new theoretical concepts 2. Low energy
bound systems 3. l~sses 1940-2500 * 4. Masses 2500-00 and
5. Concluding remarks.
1. INTRODUCTION
Recent measurements of states in BB matter are graphically

displayed in fig. 1. It shows the masses of the measured states.
The accumulation of data in the S, T and U region are due to the
known wide meson resonances. But also in this mass region there
*all units in MeV or GeV
533
534 S. NILSSON
are new results. Most results have been obtained in the last few
months. Being new they need confirmation. The statistical signi-
ficance is, however, in most cases sufficient. Taken together, the
large amount of new data indicate that something is happening. It
could be that the long expected two quark-two antiquark baryoniuffi
states at last have been discovered. We will start by explaining
what baryonium Llight be.
Let us begin with the quark structure of matter. The quark
content of mesons, strange mesons and baryons is shown in fig. 2.
While unified theory works well for the meson strings, baryons
resist a unified theoretical description. G.C. Rossi and
G. Veneziano have made progress in the possible description of
baryon dynamics in dual and gauge theories (1). One result is that
the baryon resembles a Y-shaped string. This is illustrated in
fig. 1. The graph does not represent a number or a process, it
illustrates the structure of a mathenlatical operator, the coupling
of colour. The baryon is not A- shaped or a string with three
equally interspaced quarks. It is obvious in fig. 2, that in add-
ition to the three quarks, the baryon is characterized by a junction,
where only the gluons couple. For the string junction there is a
QCD interpretation in terms of certain colour indices flowing
through the Feynman graphs (1). So in addition to the three quark
flavour lines we will draw a fourth dotted line to remember this
fact. We are not interested in all possible theoretical views of
junctions in this paper, but note that it is a useful concept for
bookkeeping, when one is dealing with baryons.
We illustrate the lowest order diagrams contributing to anti-
baryon-baryon scattering in fig. 3. The left hand side diagrams
illustrate from top to bottom the contributions from the Pomeron,
one meson (qq), two meson (2q 2q) and three meson (3q 3q) exchange.
Duality was always in trouble with BB scattering. J.L. Rosner
NARROW RESONANCES IN SB REACTIONS 535
Mass
GeVlc'
~r------------------------------'
decoy modu
3
-
U
--
T
2 _PI>
5
- - - - - - - - - - - -pp -Ihr.-~Id- - -
Fig. 1 Display of recent measurements of states in BB matter

-
(see also table 3).
536 S. NILSSON
L
u
I u
u a
d
I
Mesons
u
Strange mesons
p n
Baryons
Fig. 2 Quark structure of matter. Note Y-shaped baryon.

NARROW RESONANCES IN SS REACTIONS 537
noted the difficulties already in 1968 (2). The one meson exchange
in the t-channel was dual to a four quark structure (Zq Zq) in the
s-channel. This four quark structure has been with us for ten years
and now we are reaching the heart of the matter. What ~s this four
quark structure? Is it a physical state? Does it exist? Does it
decay? How?
In view of what was said before, we will be careful and also
draw the flov1 of the colour structure, the junction line. This ~s
done in the figure. The dotted line says "I remember I was a
baryon". By s- and t-channel symmetry we now discover, that we
have a whole new set of true dual diagrams (right hand side of
fig. 3) which also contribute. These now represent annihilation ~n
the sense of baryon number annihilation in the t-channel. In the

one meson exchange, the s-channel structure is a meson with Zq and
J
Zq and Z junction lines, which we denote by M4 J stands for junc-
tion, the lower index for the number of quarks. This object is
being exchanged in the t-channel in the dual diagram giving an
intermediate s-channel (qq) mesonic annihilation state. Going to
the other two scattering diagram Zq Zq t-channel exchange correspond
J
to a state with qq and Z junctions in the s-channel, MZ ' and fin-
ally 3q 3q - exchange to M~ (no quarks in the s-channel but two
junctions). This is illustrated in fig. 4.
At last we have defined what we mean by "baryonium". It is the
J J J . .
M4 , HZ' MO objects illustrated in fig. 4. G.C. Rossi and G. Venenano
have estimated the trajectories for the M-particles and the inter-
cepts for t = 0 are also given in the figure.
An earlier model of the baryon was suggested by D.B. Lichtenberg
(3), where the baryon structure is an antidiquark (Q) - quark (q)
pair. The diquark Q = (qq) being two antiquarks. The Qq model of
the baryon (fig. 4, bottom) is a special case of the Y-shaped baryon
with one degree of freedom less. The quark-antidiquark system was
538 S. NILSSON
Scattering Annihilation
-------------
'II
dual
..
l[
---_/ '---
]
------------
[ ][
._-- ""' .. _--
- ---------
]
------------
f[ ) [
] [ ,---
I
tI
"- -
Fig. 3 Lowest order diagrams in BB scattering

-
------ flavour line ... colour coupling line
NARROW RESONANCES IN SB REACTIONS 539
q q
juncllon Int.rc.pt
I
---------
--------- Mt :( 2q 2q) a 4 .. --12
\ no of quarks
q Cj
M~ =(q q)
<1> aO" "2

1
T:
(antldiquarkl
Qq mod.1 of baryon
Fig. 4 Baryonium systems

------ flavour line colour coupling line
Chew-Rosenzweig model of baryon (bottom).
540 S. NILSSON
recently applied to BB scattering in two papers by G.F. Chew (4)

and C. Rosenzweig (5), where also the concept of baryonium made a
break-through. Baryonium in the Chew-Rosenzweig sense is a QQ-
system. It corresponds to our previous Hi baryonium. The only
difference is that the diquark system (Q) is assumed to have baryon
number B = 1 and spin S = 1.This leads to a number of predictions
for the quantum numbers of baryonium (see concluding remarks). The
conservation of the junction lines or the diquark lines leads to
the desired property of preferred decay into a BB pair.
H.1. Hiettinen emphasized (6) already a year ago that "there
is a compelling theoretical need for exotic baryonium states". We
will now review recent experimental results, which might have some-
thing to do Hith baryonium.
2. LOW ENERGY BOUND SYSTEMS
The oeson exchanges (fig. 3) can be parametrized in terms of

a classical potential. The well-known NN potential is connected to
the
-
NN potential by G-parity conjugation. Using this as an input
the NN states can be predicted. 1.S. Shapiro (7) et al. pointed out
that the one boson exchange potential will give rise to many NN
bound states around the NN-threshold. See review by T.E.O. Ericson
(8)
The predictions for the Q-values of the pp protonium states by
L.N. Bogdanova et al. (7) and C. Dover (9) are shown in fig. 5. The
figure shows the lowest lying states below the pp threshold.
A Basel-Karlsruhe-Stockholm collaboration has measured gamma
rays from pp annihilations at rest. In addition to the continuum
the experiment finds evidence for gamma rays of energies 183, 216
and 420 UeV (10). It is the first evidence for such transitions.
Being an extremely important result it must be experimentally estab-
lished beyond doubt. I take the result at its face value and will
NARROW RESONANCES IN BB REACTIONS 541
Protonium states In potential theory

Q value
'fl,
-
----
'D,
- 'P, ---'P,
---'P,
10 0 ----'P,
----'P,
___ ~51
=1~~
_ _ _ _ 'S,
183
'So
200 216
---
420
MeV --r
Shapiro et 01 Dover ct 01
Fig. 5 Protonium (pp) states in potential theory.
------------
1897
I
3
21!..~'-'C-------'-'-.-:-+-.TT"""~
2 - prong
6 - prong
-40 -20 20 40 60
Q MeV
Fig. 6 Invariant cross section (arbitrary units) for off mass

shell p'n'-+- pions as a function of the Q-value.
542 S. NILSSON
tentatively interpret it as evidence for gamma transitions to bound

baryonium states. This is illustrated in fig. 5. The levels agree
,
approx~mate 1y w~t ' d l So an d 3Sl 1evels. The leve 1
' h t h e pred~cte
corresponding to the 420 HeV transition (mass 1455) could be due to

MOJ b aryon~um
' (pure glue) with estimated mass of 1520 (11). See
also a recent discussion by F. Myhrer (12).
Another .,ay of studying near threshold states is to make use
of off mass shell annihilation. This has been done by T. Kaloger-
-
opoulos et al. (l3) in a study of pd reactions. The NN decays to
mesons are measured in this case. If t is the lab. momentum of the
antiproton and q the spectator proton lab. momentum, the invariant
cross section on the bound neutron can be written
+ + 2
f a: 1
p + qlo p'n' (Q,q 1m)
+ + 2 2
Q H - 2m ~ (p + q) ~
pn 4m III
This quantity has been plotted as a function of Q in fig. 6. It

shows evidence for a peak at 1897 appearing in all topologies.
There are new results on the well-known 1935 meson (14, 15, 16).
It is definitely established as a narrow resonance. The best
estimates give
M 1936 1
r 4 - 8
I 0
J 2
for its parameters.

The potential theory has two candidates for this state 3D
2
with mass 1925 and lD with mass 1955 (7)
2
The troublesome point is, however, that 1935 decays to pp but
not to un (15, 17). This is illustrated in the figs. 7 and 8.
NARROW RESONANCES IN S"B REACTIONS 543
12 6"i ( ji"p)
D
e
0
b
<J -4
-8
- 12
9 ~
8
6 ,,;, (h)
..,
e
b~
<J - ,
1910 1910 1930 1940 1950
Moss Mevlc 1
Fig. 7 /::'0 = o (peak) - o (background) for pp 0 tot and 0

e1
at
the 1935 peak. Solid line, fit of ref. (18)
200 300 400 SOO 600 100 p MeVk
pp-iin
24
t
t f Ht +
o tot + t +
f 0+ ~++!
.D
E 16
b 0
0
00 0
0
8
1900 1920 1940 1970 M MeV/C'
Fig. 8 Charge exchange cross section
ref. (15) o ref. (17)

544 S. NILSSON
Fig. 7 shows~0 , which is the cross section at the peak with a

smooth background subtracted. 1935 appears both in the total and
elastic cross section. The data are from V. Chaloupka et al. (15).
The solid curve is a fit by R.L. Kelly and R.J. Phillips (lB) of
a narrow resonance H = 1934, r = 9 and an interfering background.
The condition 1935 f nn is imposed. We show the charge exchange
cross section pp + nn in fig. B.
Other explanations of the absence of 1935 + nn is to invoke the
interference of two degenerate lD2 resonances with I = 0 and 1 (19).
However, this leads to too large total widths in disagreement with
the data.
So the 1935 meson can be phenomenologically parametrized and
understood, but it is a rather artificial understanding. Let us
also point out the result of W. Bruckner et al. (16), where rs 4 MeV.
The 1935 could be very narrow.
An explanation along entirely new lines has been given by
G.C. Rossi and G. Veneziano (20). They discuss the possibility
that the electromagnetic mixing might be comparable to quark con-
figuration mixing and nearly pure quark states could result. The
1935 could be a 90 % uuuu state, which would explain the suppression
of 1935 + nn.
3. HASSES 1940-2500
Ue will begin this section with a few remarks on off mass shell
annihilation. It turns out to be the ideal tool to prepare the
initial and final states for discoveries of baryonium matter. We
show examples of triggers p + TI+ TI- + P and p + TI- in fig. 9.
This ensures that the ~++ 1S exchanged in the t-channel. With the
colour flow as in the figure, a true mesonic state will be produced
at the lower vertex.
Now suppose instead, that we have the colour flow as in fig. 10.
p
n
~]e:~~_
d ...;;...... d
d d
Tt- gJ" rem:: P
U: x-
p ~ ............. ) d
p ~]L!C:IT-
H i x+
j5 ~ ..ii ...... .. d
Fig. 9 Quark diagrams for all mass shell annihilation.
:-:)RC:: :-
d ............................... d
Fig. 10 Quark diagram for 1T - P -+PFX- illustrating possible

creation of }S baryonium.
546 S. NILSSON
We could even select candidates for this process by selecting or

triggering on decays X- -+ np, pp Tf- etc. Other charged modes
can be studied by selecting a Tf- at the upper vertex.
A CEr~ Omega spectrometer experiment has studied 9 and 12
GeV / c Tf- P -+ PF PP Tf- reactions, P. Benkheiri et al. (21) The events
were selected using a fast forward proton trigger. In this case
also a forward Tf- was selected. Fig. 11 shows the reaction. So,
in fact, the process
-+-
pfl -+ pp
was investigated. The results are shown in fig. 12. The events
have been selected by the requirement of 1175 < m(PF Tf -) < 1300 and
cosS GJ < 0, where 9 GJ is the Gottfried-Jackson angle for the proton
in the decay of U -+ pp. This reduces the background.
In addition to the 1935 peak the experiments find peaks at
11 = 2020 3 r 24 12
and M = 2204 5 r 16 + 20
- 16
The most remarkable thing is that the peaks are so narrow.
The cross sections for the two-body reactions
have been estimated, table 1.
Table 1. Cross sections for Tf- p -+ fI (N) M, nb
State 110
1935 10 5 7 5
2020 15 5 30 10
2204 20 5
9 and 12 GeV/c tt-P

P
P
tt- _ ..
Pf
1<
13 GeV/c tt-d
P n
P or n p p X
tttt
Fig. 11 Exchange diagrams for the 9 and 12 GeV/c1T-p (21) and

-
13 GeV / c 1T d (28) experiments.
30
N
~
>II
~ 20
-C
o
N
III
~
ILl
10
2000 2200 2400

P P Invariant mass MeV/c!
Fig. 12 pp mass for 1T -p+(P F "-) PP with PF 1T in 6 (1230) band

and cos e PP
< 0 for the PP decay angle.
548 S. NILSSON
The cross sections have large errors but the absence of NO (1520)
MO (2204) might be significant and might have to do with the quark
contents of the particles. In the spirit of this talk we want to
point out the possibility of an intermediate state M~
p /':,.+ -
pp.
In the final state
pp -+ - 1T+
'IT
on the contrary, the annihilation states are selected and one

would expect to find only s-channel meson states, fig. 3, right
hand side.
This is indeed so as is born out by the beautiful measurements
by E. Eisenhandler et al. (22), which have been analyzed by
A.A. Carter et al. (23). See also ref. (24) and (25). Amplitude
analysis of polarisations and differential cross sections. fig.13,
yields the following meson states (best estimates)
JPCI G +
M 2190 10 r 100 20 3 1
M 2350 15 r 200 20 4++0+ and I 1+
-- +
U 2480 30 r 280 25 5 1
We note the clear difference between the widths of the states we

are interested in in this paper, a few MeV, and the meson resonances,
a few hundred MeV.
4. MASSES 2500 - 00
I will present the masses in increasing order. This will give

rise to some unphysical jumps in the presentation, but never mind.
.1.0 - -- ------ - ---

,D o LO
o 'D
cosce") cos CS-)
~
1/1
SJ
;i.
-
c:
"C
b
"C
cos (e-) cos (e")
Fig. 13 Sample of fits for po1arisations and d 0/ d Q for pp +

n n+ in the mass region 2100 - 2500.
550 S. NILSSON
With increasing masses we quite naturally encounter increasing

difficulties and increasing uncertainties.
Next state we find is a narrow peak in the Ko 7f-7f
+ of' -
7f mass
s
+ - 0
at 2.60 GeV in 12 GeV/c pp KO 3:'rr 37f X interactions in BEBC (26).
s
We shoy, the evidence in fig. 14. The peak has mass and width
1>1 2.60:1:0.01
r < 0.018
and (j BR - 20 ]J b. BR = Branching Ratio
The peak is also found in the (A Nn 7f

-
r+ + cc effective mass.
The production process could be
pp -
+ NUX
LK37f
where M denotes a meson.

Decay of baryonium into
-QQ + MM
-
is likely to be suppressed (no conservation of diquark or junction

lines). The decay into BB is according to expectations for
baryonium.
A similar experiment in the CERN 2 m HBC (27) finds no evidence
for the peak, fig. 15. It is, however, more difficult to prove that
an effect is not there and the significance for that is 20. The
resolution in the 2 m HBC is two times the resolution in BEBC.
The conclusion is that more evidence is needed.
The same is true of a result from a streamer chamber experiment
at SLAC (28). The experiment measures 13 GeV/c7f- d + PF(PP 7f- or
np- 7f). The NN7f mass, fig. 16, shows a peak at 2670 with a width
leo
.
~
GO
- 4~-
".;
~
..
<
2 40 9
~ 40
~
! "~
C>
.... 200
t
0
0
.
....
c
20
.g
0
c 160
D
e
e O L-~~~~~~~~
2~ 2..0 26-
'0 M lie: ...... . ) CCtVIC ' ,
.
D
120
E
~
z
80
40
0
10 20 Z.:i
o + - 0 + _
Fig. 14 Hass of K 1f- 1f+ 1f- for 12 GeV/c pp +K 31f 31f X
s S
reactions in BEBC.
250 .
>OJ
::t: 200 .
,...,
0
......
150 .
..0
E
-
0
u
100 .
C>
~
.D 50 .
E
:>
Z
~~~~1.~6~~~2~.~
2----~~--~~~----~~. 0
MtK ... n-n --'
Fig. 15 Mass of KO .,.,. orr + - / 0 +

s " " 1f for 12 GeV c pp +KS 3 1f 3 1f X
reactions in the CERN 2 m HBC.
552 S. NILSSON
24 X-_ ppn-
20
..
~
>
::I: 16
on
~
.
l!I 12
c
ell 8
0
2000 2200 2UO 2600 2800
M(NNn) MtV/C 2
lJ. In and u < O. S

40
.
~
X--_pTl-n
>
::I:
0
..
2
l
20
Fig. 16 }fuss of NNn

- in 13 GeV/c n d reactions in a streamer
chamber at SLAC.
r~o. It also decays into an odd number of pions implying G = -1.

The statistics corresponds to 130 events/ ~b. The similar CERN
Omega experiment (21, second reference) finds no evidence for the
peak with 10 times more data. There are, however, differences in
acceptance and selections, which must be evaluated:
A CEP~ Omega Spectrometer experiment studying
with a fast, forward antiproton trigger finds strange mesons
H -+ fI. N
s
at 2.80 and 3.05 Gev/c 2 , fig. 17 (29). The peaks are not narrow
I'V 100 HeV. These states could be produced by exchange of baryonium,
but for the moment we just note their possible existence.
A most remarkable result comes from another CERN Omega experi-
ment with the same trigger but using TI instead
fast proton, p = slow proton

s
The (pp TI ) - mass distribution is shown ~n fig. 18. A new
resonance with the parameters
m = 2.95 0.01
r-o
a BR"'" 1 ~ b
appears (30).
If one looks at the decay modes by plotting the pp mass, fig.
19, with the events defined as the peak minus the events in the
wings, properly normalised, there is indication of decays
12 G.V/c KP-~NN(Tt) c.n
c.n
L.~f ~
130 I lin
oco 0'"
N M
110~ lill
, , L 11
~
>
..
C>
TOTA L ~p,IT ' MASS SPECT~M aT 16 ~(t " <:&62 ......,,'
.,z
q .,.", .,. a
0
-
.l!/
c p.
> ::::
~ ! ..... -
....
'0
D
t
E
;,
z
2.0 2.8 3.6 4.4 4.0

MASS (ii~7T-) GeV ~
M(AN) GeVlc l Z
Fig . 18 Uass of pp TT in 16 GeV/c r
+ - ~
Fig . 17 Bass of A N in 12 GeV/c K p-+- l\. NN ( 1T ) reaction s . 1T P -+- (pp1T - ) p .. reactions . o
z
1
PEAK-WINGS
4
10
5
~
C)
vo O~~I~~--hHH=--~-trL---
6
a:: _ ns7T+(4996 events)
~ 7/'-P-PPF7T -/ p s7TO(3341 events)
~
z
w
~ 10
~~--~~~--~--~I~~
2.0 2.4 2.8
MASS PPF' GeV
Fig. 19 Mass of p~ in the decay 2.95 (ppTI ) (pp) + TI

556 S. NILSSON
2.95 -+- 2.02 1f ?
-+- 2.20 1f
-+- 2.60 1f
But there is no indication of the decay mode
++
2.95-+-Ll p
The 2.60 - 2.65 peak is uncertain because of steeply falling wings

and corresponding uncertainty in the subtraction. The important
point is that the 2.95 might decay via other possible baryonium
states.
The production process could be exchange of baryonium
Baryonium is exchanged in pp annihilation and there is no reason

why it could not also appear in other production processes.
5. CONCLUDING REUARKS
A wealth of new data comes mainly from CERN Omega spectrometer

experiments. They are statistically well established. If confirmed.
the new resonances are a great contribution to new physics.
We will conclude with a few general remarks. An old dilemma
has been the question if the annihilation cross section equals the
difference between the pp and pp cross sections
CJ
ann
= LlCJ = CJ (pp) - CJ (pp)
The above relation can be used to obtain the distribution of

in the impact parameter space by transforming d d /dt. The result
is shown in fig. 20 (31). The difference LlCJ is very peripheral
with a peak ff'"oj 1.2 fm. On the other hand. the annihilation is
-
0.8
0.4
biN
"O:g
0
-II::
'--
<J -0.4
-0.8
0 0.5 1.0 1.5
b (frn)
20 The cross section for cr(pp) as a function of the

impact parameter b.
558 S. NILSSON
likely to be very central (32). We are faced with a contradiction.

Some new views on this problem are given in the work of G.C. Rossi
and G. Veneziano (1). They state that L.a - a
measures the
ann
importance of the w -traj ectory. In such a case a can be central,
ann
while the (;J -trajectory contributes to the peripheralism of L.a.
Still we are faced with the fact that L.a - a
ann
= 0 up to 12 GeV/c
(33).
The connection of the new resonances with baryonium is of
course an open question.
In the model of G.F. Chew and C. Rosenzweig (4, 5) the quantum
numbers of baryonium are given in table 2. S = 0, 1 or 2, P = (-l)~
and C = (_1)1+ s .
Table 2. Quantum numbers of QQ baryonium

Angular Corresponding
momentum spin J PC NN state
i, S=o J PC = 0++ 3p
0
0
1 1+- 1p
1
2 2++ 3p 3F
2 2
i 1 3S 3D
l- 0
1 1
1 0-+ is
0
1-+
2-+ 1D
2
3S 3D
2 1 1 1
3D
2
2
3D
3
3
Tentative trajectories are shown in fig. 21 (4). They are

split according to S =2 (highest) S =1 (middle) or S =0 (lowest)
and exchange degenerate in t and I. The 1935 meson is marked on
++
the S = 0 trajectory. Its quantum numbers should be 2 But
potential theory would prefer 2-- or 2-+ (7). The wide 3-- and
4++ mesons also fit well onto this trajectory.
Another estimate of the baryonium trajectories has been given
by G.C. Rossi and G. Veneziano (1). From a multiperipheral model
they obtain
1
MJ 0. 4 - - + t
4 2
MJ 0. 2 lt
2 2
1 1
UJ 0.0 + - + -t
0 2 3
J
These trajectories are also shown in the figure. The H4 trajectory
and G.F. Chew's S = 2 trajectory coincide. The gan~ now is to put
physical particles on the trajectories. We have many candidates
at our disposal but further progress can only come when we know
(i) the width of the 1935

(ii) quantum numbers of the 1935
(iii) quantum numbers of the 2020 and 2204
(iv) does the'- 2.65 particle exist?
(v) quantum numbers of the 2.95
(vi) s-channel structure of the annihilation.
The s-channel structure of annihilation should reveal the nature

of the exchanges. This is illustrated in fig.22. Corresponding
to ~, ~ or ~~ exchange, the annihilation should have a 3, 2 or 1
jet structure. The idea of annihilation proceeding through
rearrangement of a 3q 3q quark system was numerically found to be
560 S. NILSSON
i!
'"
'"
1111111
5
3
I. "I' GeV'
Fig. 21 Possible baryonium trajectories.

;0
3 q
M~ exchange
..
., 3 Ii
3 Jet slructure
"
leading I erm
2q
M~ exchange 2 jet structure
;0 q
M~ exchange I jel sir ucl ure
--- q
Fig. 22 Structure of the s-channel in annihilation.
Table 3
Summary of measurements
Decay
Mass MeV Width MeV or reaction References
1455 420 17
1660 y-rays 216 9 10
1695 183 7
1795 < 8 pd 36
1897 ~ 25 pd 13
1932 2 94 14
1933 ~5 pd 13
1936 1 94 15
1939 3 <4 16
~ 1960 ~ 80 -> 517 37, 29
1975 1 <2 -+ odd 17 38
1986 1 ~8 - KK .. 38
~ 2000 100 20 -+ 517 37, 29
2007 5 105 10 -+ KKl7 39
2020 3 24 12 21
2150 6 116 15 22
2150 30 200 25 3--1+ 23
2187 3 56 8 40, 24
2193 2 98 7 24
2204 5 16 +20 21
-16
2310 30 210 25 4++0+ 23
2335 6 132 15 22
2350 60 190 60 4++0+ 41, 29
interference
2359 2 1.65 12 24
2363 2 171 10 40, 24
-- +
2480 30 280 25 5 1 23
2600 10 < 18 -+ K 317 26
2670 ~O 28
2800 -+ AN 29
2850 pd 42
2950 10 ~O 30
3050 - AN 29
3050 pd 42
562 S. NILSSON
reasonable a long time ago by n.R. Rubinstein and H. Stern (34)

Experimental tests for exotic states using off-mass shell
annihilation was proposed by H.J. Lipkin (35) also some time ago.
The questions are tough, but the answers will surely be
rewarding. We are approaching the solution to a great problem in
quark dynamics, namely the properties of the mesonic four quark
system.
We finally give (in table 3) a list of all recent measurements,
which have been reviewed. Some have not been explicitly mentioned
in the text and we refer to the table and the references therp..
ACKNOWLEDGEMENTS
My task was made very easy by having access to the excellent talk
by L. Montanet at the Boston Conference (29) for which I am very
grateful.
REFEHENCES
1 G.C. Hossi and G. Veneziano, TH.2287-CERN (1977).

G. Veneziano, TH.2311-CERN (1977).
2 J.L. Rosner, Phys. Rev. Letters, II (1968) 950.
3 D.B. Lichtenberg, Phys. Rev. 178 (1968) 2197.
4 G.F. Chew, Proc. of the Third European Symposium on Antinuc1eon-
Nucleon Interactions, Stockholm (1976) 515 (Pergamon Press)
preprint LBL-5391 (1976).
5 C. Rosenzweig, Phys. Rev. Letters, 36 (1976) 679.
6 H.I. t1iettinen, Proc. of the Third European Symposium on }~ti
nucleon-Nucleon Interactions, Stockholm (1976) 495 (Pergamon
Press).
7 I.S. Shapiro, Soviet Phys. Usp. 16 (1973) 173.
L.N. Bogdanova, O.D. Da1karov an~I.S. Shapiro - Ann.
Phys. 84 (1974) 261.
8 T.E.O. Ericson, Proc. of the Third European Symposium on Anti-
nucleon-Nucleon Interactions, Stockholm (1976) 3 (Pergamon Press)
NARROW RESONANCES IN BS REACTIONS 563
9 C. Dover, Pr.oc. of the IV International Symposium on Nucleon-

Antinucleon Interactions, Vol. 2 (Syracuse, NY, 1975) p. VIII,
37.
J .M. Richard, M. Lacombe and R. Vinh Mau, Phys. Letters 64B
(1976) 121.
10 P. Pavlopoulos, G. Backenstoss, P. Bluem, K. Fransson, R. Guigas,
N. Hassler, H. Izycki, H. Koch, A. Nilsson, H. Poth. M. Suffert,
L. Tauscher and K. Zioutas, Evidence for Narrow Bound States
Related to the pp System, Submitted to Physics Letters B.
11 H.G. Dosch and H.G. Schmidt, TH.2296-CERN.
12 F. Hyhrer, TH.2348-CERN (1977).
O.D. Dalkarov and F. l1yhrer, TH.2280-CERN (1977).
13 T.E. Kalogeropoulos and G.S. Tzanakos, Proc. of the Third
European Symposium on Antinucleon-Nucleon Interactions,
Stockholm (1976) 29 Phys. Rev. Letters ~ (1975) 1047 .
14 A.S. Carroll, I-H. Chiang, T.F. Kycia, K.K. Li, P.O. Mazur,
D.N. Michael, P. Mockett, D.C. Rahm and R. Rubinstein, Phys.
Rev. Letters ~ (1974) 247.
15 V. Chaloupka, H. Dreverman, F. Narzano, 1. Montanet, P. Schmid,
J.R. Fry, H. Rohringer, S. Simopoulou, J. Hanton, F. Grard,
V.P. Henri, H. Johnstad, J.H. Lescetix, J.S. Skura, A. Bettini,
M. Cresti, L. Peruzzo, P. Rossi, R. Eizzarri, M. lori,
E. Castelli, C. Omero and P. Poropat, Phys. Letters 6lB (1976)
487.
16 W. Bruckner, B. Granz, D. Ingham, K. Kilian, U. Lynen,
J. Niewisch, B. Pietrzyk, B. Povh, H.G. Ritter and H. Schroder,
Phys. Letters, 67B (1977) 222.
17 ~1.Alston-Garnjost, R. Kenney, D. Pollard, R. Ross, R. Tripp
and H. Nicholson, Phys. Rev. Letters 35 (1975) 1685.
18 R.L. Kelly and R.J. Phillips, Rutherford Laboratory preprint
76-053T159, Lawrence Berkeley Laboratory preprint 4879.
19 C.B. Dover and S.II. Kahana, Phys. Letters 62B (1976) 293.
20 G.C. Rossi and G. Veneziano, TH.2342-CERN (1977).
21 P. Benkheiri, J. Boucrot, B. Bouquet, P. Briandet, B. D'Almagne,
C. Dang Vu, B. Eisenstein, A. Ferrer, P. Fleury, B. Grossetete,
G. Irwin, A. Jacholkowski, A. Lahellec, H. Nguyen, P. Petroff,
F. Richard, P. Rivet, G. De Rosny, P. Roudeau, A. Rouge,
M. Rumpf, J. Sic, J.M. Thenard, D. Treil, A. Volte, D. Yaffe,
T.P. Yiou and H. Yoshida, LAL Orsay P-1977/3.
564 S. NILSSON
J. Boucrot, B. Bouquet, B. D'Almagne, A. Ferrer, A. Jacholkowski,

A. Lahellec, P. Petroff, P. Roudeau, J. Six, F. Navach,
P. Sonderegger, D. Treille, P. Rivet, A. Volte, P. Benkheiri,
G. De Rosny, A. Rouge. R. Salmeron, J.P. Wuthrick and H. Yoshida,
Nucl. Phys. B121 (1977) 251.
22 E. Eisenhandler, W.R. Gibson, C. Hojvat, P.I.P. Kalmus,
L.C.Y. Lee, T.W. Pritchard, E.C. Usher, D.T. Williams,
H. Harrison, W.H. Range, H.A.R. Kemp, A.D. Rush, J.N. Woulds,
G.T.J. Arnison, A. Astbury, D.P. Jones and A.S.L. Parsons,
Nucl. Phys. Bl13 (1976) 1.
E. Eisenhandler, Vl.R. Gibson, C. Hojvat, P.LP. Kalmus,
L.C.Y. Lee, T.W. Pritchard, E.C. Usher, D.T. Williams,
M.A. Harrison, W.H. Range, H.A.R. Kemp, A.D. Rush, J.N. Woulds,
G.T.J. Arnison, A. Astbury, D.P. Jones and A.S.L. Parsons,
Nucl. Phys. B96 (1975) 109.
A.A. Carter, M. Coupland, E. Eisenhandler, C. Franklyn,
W.R. Gibson, C. Hojvat, D.R. Jeremiah, P.LP. Kalmus,
T.I1. Pritchard, M. Atkinson, P.J. Duke, H.A.R. Kemp,
D.T. Williams, J.N. Woulds, G.T.J. Arnison, A. Astbury, D. Hill
and D.P. Jones, Proc. of the Third European Symposium on Anti-
nucleon-Nucleon Interactions, Stockholm (1976) 117.
23 A.A. Carter, 11. Coupland, E. Eisenhandler, W.R. Gibson,
P.I.P. Kalmus, D.P. Kimber, A. Astbury and D.P. Jones, Phys.
Letters 67B (1977).
24 J. Alspector, K.J. Cohen, W.C. Harrison, B. Maglich, F. Sammes,
D. Van Harlingen, G. Cvijanovich, M. 11atin and J. Oostens,
Phys. Itev. Letters, 30 (1973).
25 E. Eisenhandler, Proc. of the Third European Symposium on Anti-
Press).
26 A. Apostolakis, R. Casali, C. Caso, Y. Goldschmidt-Clermont,
L. Pape, J.P. Porte, A. Stergiou, B. Tallini, G. Vassiliadis,
H. Henninger, G. Grard, V.P. Henri, P. Herquet, J. Kesteman,
S. Banerjee, K.W.J. Barnham, R. Beuselinck, I. Butterworth,
J.R. Campbell, J. Chaff, H.E. Mermikides, D.B. Miller,
D. Bertrand, D. Johnson, J. Lemonne, P. Renton, J. Wickens,
F. Van den Bogaert, B. Daugeras and A. Jacholkowska, Phys.
Letters 66B (1977).
27 G.W. van Apeldoorn, D. Harting, D.J. Holthuizen, B.J. Pijlgroms,
M.M.H.M. Rijssenbeek, D. Visser, J.M. Warmerdam-de Leeuw,
V. Karimaki, R. Kinnunen, }1. Korkea-aho, J. Tuominiemi,
W.H. Evans, P. Johnson, P. Mason, H. Muirhead, G. Patel,
G. Blomquist, G. Ekspong, T. Moa, S. Nilsson, Ch. Walck and
N. Yamdagni, A search for the I-Peak at 2.60 GeV in 12 GeV/c
pp Interactions in the 2 m CERN RBC, submitted to the Int. Conf.

on High Energy Physics, Budapest, July 1977.
28 A. Rodgers, Paper submitted to the Vth International Conference
on Experimental Meson Spectroscopy, Boston, 29-30 April, 1977.
29 L. t1ontanet, Experimental Review of Baryon-Antibaryon Inter-
actions, talk given at the Vth International Conference on
Experimental Heson Spectroscopy, Boston, 29-30 April, 1977.
30 C. Evangelista, B. Ghidini, A. Palano, V. Picciarelli, G. Zito,
P. Hattig, K. Huller, E. Paul, W. Ruhmer, B.R. French,
\J.A. Mitaroff, C. Palazzi-Cerrina, R. Strub, A.S. Thompson,
P.L. Woodworth, ~1. Edwards, T. Armstrong, J. Gordon, I. Hughes,
G.M. Lewis, R.~l. Turnbull, C. Best, R.A. Donald, D.N. Edwards,
~l.A. Houlden, C. Costa, L. Mandelli, S. Pensotti, L. Perini,
D.H. Miller and H.D. Lambacher, CER-N'/EP/PHYS 77-24 (1977).
31 Fermilab Single Arm Spectrometer group, see ref. 6.
32 P. Gregory, P. Hason, H. Muirhead, G. vlarren, C.J. Hamer,
G. Ekspong, R. Carlsson, S.O. Holmgren, S. Nilsson,
R. Stenbacka and Ch. Walck, Nucl. Phys. El02 (1976) 189.
33 S. Nilsson, Proc. of the Third European Symposium on Antinucleon-
Nucleon Interactions, Stockholm (1976) 365 (Pergamon Press) .
34 H.R. Rubinstein and H. Stern, Phys, Letters 21 (1966) 447.
35 H.J. Lipkin, Phys. Rev. D7 (1973) 2262.
36 L. Gray, P. Hagerty and T. Kalogeropoulos, Phys. Rev. Letters
26 (1971) 1491.
37 C. Defoix, L. Dobrzynski, P. Espigat, P. Ladron de Guevara,
M. Laloum, P. Lutz, C. Angelini, L. Bertanza, A. Bigi, R. Casali,
R. Pazzi and C. Petri, Proc. of the Third European Symposium on
Antinucleon-Nucleon Interactions, Stockholm (1976) 175.
38 A. Subramanian, Proc. of the Third European Symposium on Anti-
Press).
39 C. Angelini, L. Bertanza, A. Bigi, R. Casali and R. Pazzi, Proc.
of the Third European Symposium on Antinucleon-Nucleon Inter-
actions, Stockholm (1976) 169 (Pergamon Press).
40 R.J. Abrams, R.L. Cool, G. Giacomelli, T.F. Kycia, B.A. Leontic,
K.K. Li and D.N. Michael, Phys. Rev. Dl (1970) 1917.
41 C. Demarzo, L. Guerriero, F. Posa, E. Vaccari, F. Waldner,
R. Dulude, R.E. Lanou, Jr., J.T. Massimo, R.K. Thornton,
D.S. Barton, M. Marx, B.A. Nelson, D.C. Peaslee and L. Rosenson,
Proc. of the Third European Symposium on Antinucleon-Nucleon
566 S. NILSSON
Interactions, Stockholm (1976) 139 (Pergamon Press).

42 H. Braun, D. Brick, A. Fridman, J-P. Gerber, P. Juillot,
G. Haurer, A. Uichalon, M-E. Michalon-Mentzer and
c. Voltolini, Phys. Letters 60B (1976) 481.
DIS C U S S ION
CHAIRMAN: S. Nilsson
Scientific Secretaries: M. Dine, T. Uematsu
DISCUSSION No.1
BIZZARRI:
I shall comment on the absence of a resonant signal

for the S(1936) in the charge exchange reaction pp-tonn.
For a resonance of spin J, definite I-spin and non-inter-
fering with background, the resonant cross sections are
6t= Px total
Gi= Px(l-x) reaction or inelastic
'i. = Px~ elastic
~= Px charge exchange
having defined the elasticity x= r,:p/~~ and the pro-
portionality factor being
Z. 2-
P ="It"~ ( 2J + 1) / (f. + 1 )
(at the resonance mass m=m R,

'Z.
For the S - meson, at m=mR =1936, one has 'W~ =22 mb,
and the measured cross sections are
C;; -y 10 mb
Gt -==- 7 mb
U; -::0" 0
The problem is: How can it be that v.G""e.? I shall

discuss two possible answers:
a) There are two degenerate resonances with 1=0
567
568 DISCUSSION
and I=I, and elasticities x~ and x~. We then have

G; = P(x,. + X2.\..
ve = P (X,. + xI- )
G; = P(x~- x~)'Z.
and if x I =t xl. we get G;~ O. Wi th this assumption however
x" + xl. = tre/()" ~ 7 /10= .7
and 2J+l = r;;/("Ir~'l..(x~+ xz.)~10/(22,-.7)=.6
J ~ -.2 i.e. J = 0
The resonance must have spin 0 and therefore be prod-
uced by S or P wave pp. It is very unreasonable to ass-
ume that at a mass m = 1936 MeV, corresponding to an incid
ent momentum of 500 MeV/c, there is not a sizable S or P
wave non-resonating background. We look therefore for an
alternative possibility.
b) Resonance interfering with non-resonant back-
ground. Assuming this non-resonant background amplitude
to be mostly imaginary, we get for the observed peaks
C;; = Px t..
be= P(x + 2~x)
u.= P(x + 2Ax)
with A= imaginary part of the elastic coherent non-reson-
ating amplitude, A= imaginary part of the charge exchange
coherent non-resonating amplitude. In terms of NN I=l,O
amplitudes it is A=(A~+ Ao)/2 and A=(A~- Ao )/2. From the
small value of the charge exchange pp cross section away
from the resonance, one expects A~~AQ and therefore A A.
As for x, its value depends on the resonance spin J:
x= ~/f?r",'L(2J+l)J = .45 ~ff J=O
.15 . . . J=l
.09 if J=2
.064 if J=3
We can then have a situation, if the resonance is a
P or D wave pp state and has spin J=2 or 3, in which 2A~1,
~A, x2A. In such a situation it would follow natur-
ally
be.l
~ 2Ax c-;
This is basically the explanation proposed by
Chelapke et al. and elaborated by Kelly and Phillips,
which is a natural one (in the sense that is does not
require a very special artificial cancellation, both terms
DISCUSSION 569
contributing t0Uo being small) at leas! for certain values

of Land J.
c) I shall just mention that an alternative "natur-
al" explanation has been recently proposed by Veneziano:
and very near degeneracy of the expected resonant states
in this region, and the increasing electromagnetic split-
ting make the physical states a mixture of different I
spin states thus yielding different couplings for pp and
nn.
BERLAD:
The presence of baryonium resonances below threshold

has several consequences. First, you expect that pn
resonances will appear in pD scattering at low energies.
If these states do not decay strongly, you will see corr-
esponding tracks. If these objects do have mesonic de-
cays, you should see mesonic decays below threshold. Such
decays would seem to be in conflict with the whole idea
of baryonium, however.
BIZZARRI:
A pn resonance below threshold has been observed by

Kalogeropoulos et al. It shows up as a peak in the mom-
entum distribution of spectator protons P s in the react-
ion pD~Ps + mesons. Baryonium can certainly decay into
mesons by strong interactions. This decay mode, however,
is suppressed compared to decays to NN, in a manner simi-
lar to that of >-+,!lr vs 4>~KK.
CHEN:
In theR experiment the signal to background ratio

is less than in the pp elastic data. The St experiment
makes a cos~ cut to increase this ratio. I think that
one should make a cos~<O cut in the pp data as well in
order to get rid of the diffractive contribution to the
1936 MeV resonance.
PEREZ-Y-JORBA:
I believe that the cut in Jackson angle in theR pp

experiment was meant to make sure that the pp state was
570 DISCUSSION
produced at the proton vertex rather than the ~ one.
NILSSON:
The pp elastic data cannot be treated in this way.

The signal is too small.
ROLANDI:
Can you explain what a junction is?
NILSSON:
Baryonium is formed when a baryon (qqq) and an anti-

baryon (qqq) come together and a quark in one and an
antiquark in the other annihilate. One must explain why
the remaining system does not decay into two mesons. The
junction is a color operator with a tree-like structure.
The middle of the tree is a junction. It is preserved
and behaves like the flavor lines in a planar diagram.
SANDA:
I have heard this kind of picture, where I believe

the junction was called the soul, described some time
ago by Professor Nambu.
ANTREASYAN:
You have said that junctions are conserved, yet they

are created. This seems unsymmetrical.
NILSSON:
They can be created and destroyed, just like quark

lines.
UEMATSU:
I would like to point out that the concept of junct-

ion in the context of quark diagrams was first introduced
by the research group of Kyushu University in Japan.
DISCUSSION 571
DE LA TORRE:
These string diagrams should just be interpreted as

bookkeeping devices, since only a subset of the solutions
of the three string problem has been obtained, and these
solutions are plagued with difficulties related to a non-
closing commutator algebra. Thus it is impossible to
calculate amplitudes; the diagrams provide some insight
into these processes.
BALDINI-CELLO:
I have experi~ental data to add to that shown this

morning. At the e e- Frascati storage ring Adone a re-
sonance has been found at 1.5 GeV with a width of order
2 MeV. Preliminary data have been presented at the Buda-
pest conference. The e+e- width is of order 0.05 keV,
so any interpretation in terms of a recurrence is diffi-
cult. It is a natural J"'PC = 1-- candidate for baryonium
near threshold. Do you know of any predictions for the
e+e- width of baryonium resonances?
NILSSON:
I don't know of any specific predictions. This re-

sult, however, is quite exciting, and bears a close re-
semblance to the 1.5 GeV state suggested by the nuclear
gamma ray experiment and the estimated mass of the no-
quark baryonium M';T .
D
CELMASTER:
Why can't this state observed in e+e- annihilation

be interpreted as the bound state of a new quark?
BALDINI:
Such an interpretation is viable. But the width of

this state to e+e- annihilation is small compared to the
w, CJ and +.
e+e- annihilation width of other vector mesons, the J~
If this resonance was a vector meson, we ~
would expect a width of about 1 keY.
572 DISCUSSION
MANDELLI:
Do any of the existing models explain the widths of
these states?
NILSSON:
No.
PREPARATA:
I am disturbed that these baryonium ideas have so
little quantitative basis. I wonder also where are the
excitations which correspond to the new degree of free-
dom which the junction represents.
CELMASTER:
You showed graphs of the reaction pp ~ i + X. You
have interpreted the three gamma bumps as a glueball,
2 quark state and a four quark state. Why do you adopt
this interpretation and not that, for example, of a glue-
ball and its excitations?
NILSSON:
Certainly one could consider other possibilities.
CELMASTER:
Why don't you see this glueball in J 1'1'-+'( + glueball?
NILSSON:
One needs to know the signature for such a process
in order to look for it.
FIELD:
I want to make a remark on the single gamma observa-
tions and the corresponding bound pp states. Having seen
the data, I feel that its statistical significance is in
doubt and that we should wait for better data before
taking these states seriously.
DISCUSSION 573
NILSSON:
I feel that it is unlikely that these results are

an accident.
JUNG:
The people who did the experiment have told me that
the only conclusion they can draw is to go and do the
experiment better, with a second NaI counter to reduce
background and systematic effects.
MIETTINEN:
I would like to make a comment. For ten years,
several theorists have emphasized that duality and dual
models don't make sense if states consisting of two
quarks and two antiquarks don't exist. Rosner, Shapiro
and collaborators, the Japanese people, and Dr. Lipkin
have long advocated searches for these states. The
reasons for the rapid progress being made now are the
new experimental facilities, such as theJl spectrometer,
and the renewed experimental interest in spectroscopy
generated by the theoretical and experimental successes
of the charmonium spectroscopy.
DISCUSSION No.2 <Scientific Secretaries: T. Uematsu

and B.G. Weeks}
BERLAD:
First a remark: In the experiment Tf-p-+{FPlt )+ ....
where baryonium is produced in the forward direction, the
baryonium mass 2.67 is at the highest value of the back-
ground, and this therefore is perhaps not the appropriate
experiment for a baryonium search. Now my question: if
the 2.95 can decay into lower baryonium states with large
Q values, why is the width so narrow? Does this indicate
the violation of some new rule?
NILSSON:
We really cannot say at this time. Of course, it

may be that there is a new quantum number which forbids
the decay, but we do not know.
574 DISCUSSION
DEO:
What are the predictions of Chew and Veneziano

regarding the widths of the baryonium states?
NILSSON:
There are no theoretical predictions for the widths.
DEO:
How could the relations o(~ .. ~ i D(~: K. i~;=~ be

deduced and not the widths?
NILSSON:
The trajectories come from multiperipheral models
for these objects. For details I would refer you to the
original papers of Rossi and Veneziano.
BERLAD:
A follow-up question: we are told that in dual models
the slopes of trajectories are universal unless loop in-
tegrations are involved. Is there someone in the audience
who can ~ell us how the low values ~~ y~ and oC~ ~ ';'~
were derl.ved?
NILSSON:
I must again refer you to the original papers.
CHEN:
Is there any theoretical explanation that the 2.02
GeV particle is ~roduced strongly in association with
&1230) and the N (1520), while the 2.2 GeV is in associa-
tion with ~(1230) but not with the N~(1520)? Could the
latter be due to phase space?
NILSSON:
It might be due to the production mechanism.
DISCUSSION 575
ZICHICHI:
What is the pion energy in this reaction?
NILSSON:
9 and 12 GeV/c, added together.
ZICHICHI:
Then it cannot be phase space. It might be some-

thing having to do with the differing isospins.
Some time back there was a comprehensive examination
of the isospin Clebsch-Gordan coefficients involved in
these reactions, and the results were claimed to be in
agreement with experiment.
COMBER:
In the controversial 2." (;;, .".-) peak seen in the

16 GeV SLAC experiment, is there a combinatorial problem
to be taken into account as with the I particle seen in
the BEBC 12 GeV pp experiment?
NILSSON:
There is no problem: the plots shown include only
exclusive NN channels.
COMBER:
What then is the meaning of the "('r~..,rTr" TtTf+)('Tf'~rrTl)'\
which follows the (pplr) on the diagram you showed?
NILSSON:
Those reactions were used to get the G parity of

the state.
WILQUET:
I want only to inform everyone that the 5~ peak ob-
served in the mass spectrum at 2.6 GeV in the
576 DISCUSSION
pp experiment at BEBC has also been seen in the A~

channel, with similar statistical significance. However
the data are still being analyzed and this is very
preliminary information.
LIPKIN:
In the reaction 7r N -i> I>J-X , i f we take X to have

isospin two, then there is a unique s channel isospin
I = 3/2, since the initial state has I = 1/2 or 3/2 and
the final state has I = 3/2 or 5/2. Therefore the cross
sections for Ji"-r
-;> p~ X- and If-'''' ~ r~x-- are related by
isospin Clebsches. Do the data agree with the isospin
predictions?
NILSSON:
As far as I know, it was not reported at the Boston

conference.
MANDELLI:
Privately Rogers says that he has done the calcul-

ation and that the data agree with the isospin predict-
ions.
ZICHICHI:
If there are no more questions, I will invite

Dr. Miettinen to give us a five minute lecture, in which
he should explain to us what is a junction.
MIETTINEN:
In five minutes you cannot say very much, so let me

do this on a very elementary and concrete way. Consider
first the collision of two mesons. Picture a meson as a
bound system of a quark and an anti-quark, joined togeth-
er with some sort of gluonic string: q .rtQiK> q. When
the two particles collide, the quark belonging to one of
them may annihilate the anti-quark belonging to the other
one, and the two mesons form a resonance, the ~-meson,
say. This resonance decays back to two mesons by breaking
the gluonic string. Thus the process goes as fOllows:
DISCUSSION 577
-
~~ ~
'j
:i
't
~f
~
~
1r T rr - ---J;> 'SO ---? 'Jt'+1J'
Now, in the string model, the baryon consists of three
quarks and glue-strings as follows:
The collision of a baryon and an anti-baryon may proceed

as follows:
The intermediate "baryonium" state may decay in a similar

way as the meson decayed in our previous example: by cut-
ting the gluon-string and polarizing a new qq-pair. It
could also decay to two (or more) mesons. However, in
order to do that it would have to reshuffle its internal
structure, and that may be difficult and suppress the
decay to mesons.
578 DISCUSSION
NILSSON:
You haven't said a single word yet about what is a

junction.
MIETTINEN:
I am doing my best. In five minutes you cannot be

too formal. In the string model of hadrons you can think
of a junction very literally as the point where the three
strings meet. In the color field theory model of
Veneziano and Rossi a junction has much more symbolic
meaning.
Let me add a few more comments. If the two hadron
wave-functions have a good overlap, we should also consid-
er processes where ~ or three qq annihilations take
place:
A
Y
--;:>
CD
Veneziano and Rossi call these baryonium states M2 and
MO, respectively. Personally, I think that the theoret-
ical understanding of these low-lying "baryonium" states
is much less advanced than that of the qqqq states. The
M2 state looks to me very much as the one-loop correction
term to the normal meson wave-function:
.
DISCUSSION 579
,
, "
Meson= + .... + high spin
o ,
' ..
Finally, let me mention that Veneziano and Rossi

have written a new paper (CERN TH prepint, July 1977)
in which they discuss the electromagnetic mixing of the
"baryonium" states. Their key point is, that the elect-
romagnetic mixing effects can be so strong that isospin
is not a good quantum number for the "baryonium" states.
They suggest that instead of being diagonal in the isospin
basis, the "baryonium" states should be approximately
diagonal in the suark basis. Thus, the S(1935) is a
rather clean uuuu state, and this explains the (experi-
mentally observed) absence of the process:
S(1935) ~ ~ -
""-
Veneziano and Rossi make also many other interesting

predictions.
PARTON DISTRIBUTIONS AND THEIR Q2 DEPENDENCE
*)
N. Cabibbo
Laboratoire de Physique Theorique et Hautes Energies,
Universite Pierre et Marie Curie, Paris
1. INTRODUCTION AND OUTLINE
The facts about the parton model are, by now, standard know-
ledge.
I will briefly recall these facts and outline a picture

evolved by our group in Rome -- of the parton distribution inside
I-It )
a nucleon and some of its physical implications.
I will then describe a simple approach to the computation of

deviations from scale symmetry in deep inelastic phenomena S ). Al-
though fully equivalent to the usual one, based on the computation
of momenta of structure functions, this approach is physically more
transparent and more useful in numerical computations of the Q2 de-
pendence of structure functions. Although the basic idea can be
traced to a paper by Kogut and Susskind, the main credit for its
development into a useful tool goes to G. Parisi and G. Altarelli.
*) On leave from Universita di Roma, Italy.
581
582 N. CABIBBO
2. THE PARTON MODEL
Consider a typical deep inelastic process, such as v-N inter-

actions at high energy. Let is be the total c.m. energy. In the
c.m. system:
5 = (1)
E being the energy of either V or P (we assume E ~). Also q

\1
is the momentum transfer, P the momentum of the incident proton,
\1
and Q2 _q2.
The parton model starts from the consideration that the weak
interaction takes place on a distance r ~ 1/1Q2 which is small with
respect to the size of the proton, and on a time interval T ~ l/iV,
where V = (pq) is a variable which, in the lab. system, reduces to
the energy transfer from lepton to hadron. These conditions sug-
gest that we are really looking at the elementary constituents of
the proton on a time scale which is short with respect to the time
scale of the proton dynamics. According to the parton picture, the
interaction goes in three steps. The first contains all the time
up to the interaction: during this time, due to its interval dy-
namics, the proton slowly moves from one configuration of internal
constituents to the next with no forewarning of the coming inter-
action. Then comes the interaction which depends only on the in-
stantaneous configuration of internal constituents and the known
properties of weak (or electromagnetic) interactions. After this,
there is a long aftermath during which the elementary constituents
rearrange themselves into the final particles.
If we are only interested in the inclusive differential cross-

section, i.e., we do not care about what happens to the hadronic
debris, all that counts is the probability of different configura-
tions for the proton constituents. Actually, even less is
PARTON DISTRIBUTIONS AND THEIR Q2 DEPENDENCE 583
sufficient: since Q2 is large, the scale of distance will be such

that the interaction will take place on a single constituent. All
we need for inclusive differential cross-sections are the proba-
bility distributions for finding a single constituent with given
properties.
The beauty of the parton model is in that the parton distri-

bution is the same for different kinds of deep inelastic processes.
Once derived by one set of data, the distribution will be the same
for other sets of data, so one can make cross checks. In his lec-
tures, Miettinen discussed the applicability of the parton model
to purely hadronic, high PT processes. Can the parton model be
justified in a field theory? Drell has shown that it can if one
imposes an arbitrary cut on transverse momentum of particles. This
assumption leads to scaling parton distributions. With the usual
variables x and PT we will define combined distributions:
LX, b )
IT
(2)
If PT is limited by a fixed cut-off, it will become negligible at

high energy so that all interesting quantities will be expressed
in terms of the x distribution of partons:
4i (x) = (3)
The results of asymptotically free theories essentially con-

firm the parton model with an important modification. The parton
idea is applicable in these theories, but parton distributions have
a weak, computable dependence on Q2, i.e., one does not have exact
scaling.
584 N. CABIBBO
3. PARTONS IN A NUCLEON MADE OF

THREE CONSTITUENT QUARKS
In the last few years, our group has evolved a picture of the
parton distribution within a nucleon, based on the following idea:
we assume that the nucleon can, even at large momentum, be thought
of as made of three quarks, which carryall of the momentum. These
will be called "constituent" quarks. The nucleon will then be de-
scribed, as in the naive parton model, by a wave function that de-
. bl es (l.
pen d s on t h e space varl.a l. l..
x, PT' = I .,
2 3) 0 f t h e tree
h
quarks, as well as on spin and isospin indices. We will have:
(4)
L P~ = 0
(5)
From the wave function we can compute, through a simple inte-

gration, the x distributions for constituents: in a proton we will
thus have two functions:
(6)
These will satisfy three sum rules which express the fact that in
a proton there are two p quarks and one n quark, and they carry
all the energy:
J }=>o (x) 01 X :- 2 (7)
J 1'\..0 (x) dx ; 1 (8)
(9)
We then imagine that each constituent is seen in deep inelastic

processes -- as made of partons. The parton distributions within
a constituent are described by a set of functions
(10)
where a denotes a parton, either a quark (p, p, n, n, ) or a

gluon g, and b denotes a constituent p or n in the case of a nu-
cleon.
The parton distribution can then be obtained by convoluting

the ~'s with the constituent distributions; we have, for example,
(11)
and similarly for other parton distributions.
In this picture we need to specify:
1) a constituent wave function which determines pO(x), nO(x);

2) a set of parton distributions within constituents ~ab.
One of the more striking results of deep inelastic ep scattering

is the x dependence of the ratio F 2 (en)/F 2 (ep): recall that the
structure functions of proton and neutron can be written as:
(12)
586 N. CABIBBO
where we used the fact that [p(x)l

...!Neutron
= [n(x)]p roton = n(x) and
similarly for p. n. n, while A and A distributions are the same in
a proton and in a neutron. In a naive three-quark model, we would
e~pect p(x) = 2n(x) , p = n = A X = 0, Fen/F ep 2/3. The values
for this ratio are found to range from 1 near x o to 1/4 near
x = 1. The value observed near x o can be understood ~n terms
of a predominance in this region of a sea of qq pairs. The devia-
tion near x =1 indicates that p(x) > 2n(x) in this region, since
p, n are known from neutrino experiments to be present only at
x ~ O. In our picture we could attribute this either to
(near x 1)
or to the structure functions of constituent quarks, or to both.

It can be shown, using the convolution equations (7)-(9) that the
second alternative leads to the lower bound:
(13)
which is clearly violated by the data near x = 1.

We then choose the first alternative which implies that p-
type constituent quarks and n-type constituent quarks have differ-
ent distributions in a proton. We recall that in the simple quark
model, which incorporates an SU(6) symmetry, the nucleon is made
of three quarks in a state which is completely symmetric in both
the space-spin and the isospin coordinates. In such a scheme the
distribution of p and n constituents would be the same. We need
to break this symmetry by mixing the basic symmetric configuration
with one which is not symmetric in isospin variables and, therefore,
given the para-Fermi character of coloured quarks, also not sym-
metric in spin-space coordinates.
Here, I will not present in any detail our ansatz for the
wave function of constituents in a proton, which is discussed in
Ref. (I), but will only recall that it contains two adjustable
parameters: S which fixes the x distribution and b, a parameter
which describes the mixing of configurations of different symmetry
and gives rise to different shapes for the Po(x) and no(x) distri-
butions. The b parameter is used to fit the x behaviour of
F2 (en)/F 2 (ep), especially near x = 1.
4. PARTONS IN CONSTITUENTS
QCD suggests that the parton distribution in constituents

arises from successive emission of gluons and of qq pairs:
9
a q
9
This picture can actually be used to derive quantitative predic-

tions on the (scale violating) Q2 dependence of parton distribu-
tions and structure functions. We can think of the above diagram
as ordered in Q2: on the left we are at small Q2, and the con-
stituent is seen as a simple particle. This is the description
used in the simple quark model, adapted to the study of static
or low-energy properties of the hadron. Moving to the right, we
go to higher Q2, and the constituent lS seen as a cluster of par-
tons of increasing complexity.
We shall come back to the quantitative study of the Q2 depen-

dence of parton distributions. At present, we note how the above
588 No CABIBBO
picture suggests the general form that the ab(x) will have at some
large Q2. We see from the above diagram that the constituent a
will contain one parton with the same quantum numbers. the "valence
quark". plus a number of gluons and qq pairs. the "sea". We will
then have, if b is a quark:
(14)
The first term is the "valence" contribution and the second is the
sea contribution. In this expression, we have assumed that the
sea distribution is the same for all qq. This assumption can
easily be released by specifying different distributions for non-
strange, strange, and charm quarks. The gluon distribution will
be described by a third function:
(15)
Although the gluon distribution does not enter the expressions for
deep inelastic phenomena. it plays an important role in determining
the Q2 dependence. It can also playa direct role in applications
of the parton model to hadronic processes.
In applying our model to deep inelastic phenomena we propose

an ansatz for the functions v, s and g, which involves a few ad-
justable parameters. This ansatz is based on assuming power be-
haviours near x = 0 and near x = 1. The power near x = 0 is fixed
b y Regge cons~'d erat~ons 1.7) whOI
0
~ e t h at f
or xl= '~s use d as an a d-
justable parameter. The normalization of the three functions re-
quires one single adjustable parameter. since we have two sum
rules:
=
where F is the number of flavours present in the sea. The first

sum rule expresses the existence of one (and only one) valence
parton, the second expresses momentum conservation. We thus have
four adjustable parameters which determine the parton distribution
in a constituent quark. In our previous work we assumed a scaling
form for the 's at large Q2, so that g was not needed and we
had only three adjustable parameters. In the work now in progress 9 )
v s, g at a fixed Q2, and
we make a four-parameter ansatz for ,
compute explicitly the Q2 dependence of these functions.
5. THE SCOPE OF THE MODEL
The model outlined above for the parton distribution in a

nucleon can be used directly to study:
+ -
1) Deep inelastic phenomena, including p+p + 1 +1
Furthermore, the wave functions for constituents can be used to

study "static" properties of the nucleon. We have considered two
applications of this type:
2a) Corrections to the SU(6) prediction 1) GA/Gv = 3/2.

+
2b) Calculation of D/F for SU(3) breaking in the (,'2) baryon
octet '+)
Both computations are based on the presence of a configuration mix-

ing, discussed in Section 2. The model can then be extended to
the parton distribution in other hadrons. These are related to
those in a proton, since we expect the ' s to be the same in all
hadrons. To determine, e.g., the structure functions of a pion
we will only need an ansatz for the wave function of the pion and
3)
this might involve a single adjustable parameter Thus extended,
the model can be used to predict:
590 N. CABIBBO
3) Deep inelastic processes involving pions, in particular 3 )

TI + N -+ Q:" + Q,- +
The pion wave function can also be used directly to describe:
4) Static properties of the pseudoscalar mesons.
As an example of this, we may quote a relation between SU(3) break-

..
Lng Ln mesons an d baryons If)
The model appears therefore to be an economical one in the

sense that a large body of data can be interpreted on the basis of
a few adjustable parameters.
6. THE Q2 DEPENDENCE OF PARTON DISTRIBUTIONS
In many ways QCD is similar to QED. The Q2 dependence of

structure functions predicted in QCD has a strict correspondence
with a phenomenon well known in QED: the appearance of Q,n(E/m)
terms in radiative corrections. Since the two effects have the
same field theoretical origin. I will discuss here the case of
QED, for which many have formed some intuitive grasp. We will de-
rive a set of equations which are the QED counterpart of the
Altarelli-Parisi equations in QCD, and refer to their papers) for
the derivation of the latter.
QCD and QED differ in one important way: the first LS asymp-
totically free, while the second is not. This will have no effect
on the structure of the equations obeyed by parton distributions,
but only on the Q2 dependence of the effective coupling constant
a(Q2) in the two cases and, therefore, on the asymptotic behaviour
of the distributions.
The parton picture Ln QED has a long history dating from the
Fermi-Weizsacker-Williams approximation scheme. Applied to brems-
strahlung this can be written as:

(16)
where y(x) is the "probability of finding a photon with fraction

x of the electron momentum" and oR is the cross-section for elec-
tron scattering. We see that y(x) has the role of a parton dis-
tribution. We can compute y(x) at order a by applying the old-
fashioned time independent perturbation theory, The relevant dia-
gram is:
e-----jC"
It is convenient to calculate a combined distribution in x and PT

for which one finds
2
-t ... (i-X)
(17)
and obtain y(x) by integration over PT' The result is found to de-
pend both on x and on p~ax ~ E, We will then write:
""o.x
r (x It:) = Jf 012.fT ( (x p)I =
(18)
= (1+ (i_)()2.)t
where we have defined

592 N. CABIBBO
t = ~ (19)
We see from this example that the logarithmic dependence upon E

derives from the integration needed to pass from a multidimensional
(x, PT) to a one-dimensional distribution.
The parton distribution y(x) can be computed at higher orders

in a by considering diagrams
/// .
etc .

Each of these diagrams defines directly a combined distribution

pD(x., pi) where i denotes the particles which appear in the final
1 T
state for each diagram. These functions will be rational functions
of their arguments. The function y(x,t) can then be obtained by
integrating each distribution over all variables but one, and sum-
ming the results.
The resulting y(x,t) will now depend on t, i.e., it will con-

tain terms of the type nn(E2/m2), arising from the integration.
We can, by the same procedure, define and calculate distributions
e (x,t) for electrons and positrons. What is their t dependence?
Assume we change E by an infinitesimal: E + E + dE(t + t + dt).
This is equivalent to increasing the "phase space" of the integra-
tion by giving an (infinitesimal) increase to the phase space of
each parton. The change in y(x,t) can be described by diagrams of

the type
.2
,et c.
where the dotted lines correspond to particles in the newly avail-

able phase space (diagrams where two particles are in the newly
available phase space are of order (dt)2, and can be neglected).
We can interpret these diagrams as representing the emission

of a photon by one of the electrons already present in the parton
distribution at t. This will be an order a effect, and it must be
proportional to the sum of the electron and positron distribu-
tions e(x,t) = e - (x,t) +
+ e (x,t). This leads to
--0(
21L
(20)
We note that P is expected to be independent of a and t. In

ey
fact:
i) the extra emission is clearly an effect of order a and we have
an explicit factor a;
ii) the extra emission arises from an integration over an infini-
tesimal phase space, so it will not give size to an explicit
t dependence.
594 N. CABIBBO
In order to determine P
,we apply (20) at lowest order in
ey
a by using our previous computation in Eq. (18) which gives, at
order a, the y distribution corresponding to an electron distribu-
tion:
(21)
(i.e., an electron of energy E).
Substituting Eq. (18) on the left-hand side of Eq. (20) and

Eq. (21) on the right-hand side, we find, after some simplifica-
tion:
(22)
We can now extend the argument by applying it to the functions

+
e-(x,t). We are led to propose the following equations:
(23)
The first term is the counterpart of that in Eq. (20): it corres-

+
ponds to the process e- + e + y. When one emits a photon of frac-
tion x, one also emits an electron of fraction I-x while the ori-
ginal electron with x =1 disappears. We thus expect P
ee
to have
the form:
(24)
C can be determined by requiring charge conservation:

(25)
This leads to the sum rule
J
~
o Pee (>l) d)( = 0

(26)
i.e., using Eqs. (24) and (22)
C :" J" i-X

ciK (-1+)(
2
) (27)
o
This is infinite so we have to be more careful. A possible pro-

cedure is to introduce a cut-off, i.e., to write
(28)
where
(29)
has been chosen to obey Eq. (26) for any A. We then substitute
(28) and (29) in Eq. (23) and obtain a result which is finite as
A -+ 0:
J ~~ Pe~ tf)e(~,
~
t)
~
=.1 .~~ (1+(~t)(e~(1,tl-j e (x/l)

t
(30)
596 N. CABIBBO
To complete the program in QED we need the function P We can

ye
obtain P by noting that if we put
ye
(31)

(a photon of energy E) we can compute directly e , at order a,
through the diagram
This gives
(32)
If we substitute Eqs. (31) and (32) in Eq. (23), and compare the
order a terms, we find:
(33)
The above analysis is partially confirmed by more refined arguments

based on the renormalization group with one important modification:
the substitution of the coupling constant a with a(t) the "ef-
fective coupling constant at distance liE". The correct equations
are then
.(1:)
(34)
1
:~) J~Y (~e (~)

>l
eZ (~,t). P(~)r (~,t)) (35)
~(t) is determined by the equation:
(36)
where B is the Callan-Symanzyk function.
Equations (34)-(36), together with (21), (30) and (33), de-

termine the t dependence of the parton distributions in QED.
Note that QED is not asymptotically free: ~(t) increases

with t. Equation (36) can be directly integrated neglecting higher
order terms to give
it):
(37)
We see that ~(t) increases with t, and goes to infinity at t = 3TI/~.

This is not of direct practical interest if one wishes to use the
above equations, since the pole in Eq. (37) corresponds to E2 ~
m2 exp [3TI x l37J, a very high energy indeed. (On the theoretical
side, the higher terms in Eq. (36) can be neglected only when
~(t) < 1, so that the prediction of a singularity in Eq. (37) is
not reliable.)
In QCD Altarelli and Parisi have derived a set of equations

similar to (34)-(35) to calculate the t dependence of parton dis-
tributions, where t is now interpreted as t = In Q2/~2, ~ being a
fixed but arbitrary renormalization point. Since QCD is asymptoti-
cally free, the effective coupling constant ~(t) tends to zero when
t + 00, so that the equations become more and more reliable as we
move to large Q2.
598 N. CABIBBO
REFERENCES
1) G. A1tare11i, N. Cabibbo, L. Maiani and R. Petronzio, Nuclear

Phys. ~, 531 (1974).
2) G. A1tare11i, N. Cabibbo. L. Maiani and R. Petronzio, Phys.

Letters 48B, 435 (1974).
3) G. A1tare11i, N. Cabibbo, L. Maiani and R. Petronzio. Nuclear

Phys. B92. 413 (1975).
4) N. Cabibbo and M. Testa, Phys. Letters 21!, 96 (1974).
5) J. Kogut and L. Susskind, Phys. Rev. D9, 697 (1974) and ~

3391 (1974).
G. Parisi in Proceedings of the XI Rencontre de Moriond on
Weak Interactions and Neutrino Physics, Ed. J. Tran Thanh
Van (1976).
6) G. A1tare11i and G. Parisi. Nuclear Phys. B126. 298 (1977).
7) R.P. Feynman. "Photon-Hadron Interactions", Benjamin, New York,

(1972) .
8) For a review, see the article by S.D. Drell in "Subnuc1ear

Phenomena", Ed. A. Zichichi, Academic Press (1970).
9) N. Cabibbo and R. Petronzio, CERN preprint TH.2440 (1978).

DIS C U S S ION
CHAIRMAN: Prof. N. Cabibbo
Scientific Secretaries: L. Baulieu, R. Fischer
BACE:
What went into the derivation of~ = - .3? I think a part of

the input you said was the ratio 'i~Fae~~,~as x~ 1-
I have reason to believe that this in fact is not the case. So if
your F
we change the d ata wh at lS D .
gOlng to be ?. 5tamatescu and I
have reanalyzed the deuteron data by using a different procedure
in "solving" the integral equations for the Fermi motion and the
value changes to ~ which is different from the value of
3
i obtained
by Atwood. By the way 7 is also a theoretical prediction of Farrar
and Jackson.
599
600 DISCUSSION
ZICHICHI:
This is one of the WHY'S of this school:

x ~ 1. You claim that deuteron corrections have changed dras tically
the data. But this is an old story. Deuteron corrections are not
agreed by many authors.
CABIBBO:
Actually we have never succee d e d ~n

f ~tt~ng 1 4 ; we used a
slightly higher value because 4 ~s
1 .
an absolute
.
l~mit. I would be
interested in having this new data since we are reanalyzing the
whole model.
BACE:
You developed a picture of valence quarks which produce by

bremsstrahlung additional parton~and the higher you go in pertur-
bation theory the more non-valence type partons there will be. At
one point you replaced the coupling constant~ by the running coup-
ling cons tant D( (Q2) In QED if you let Q2 -+ 0 then 0( (Q2) should
go to 0; so in this limit your picture contain~ just the electron
2
(valence contribution so to speak). For QCD I would take Q -. 00
and will get oQ 2 )~O. In your picture wouldn't I be led to
valence quarks again whereas I know that in this case the structure
function has large contributions near x = 0 which means lots of
sea quarks ?
CABIBBO:
2
In QCD when Q ~oo you also tend to sea quarks only; you do
it very slowly:
This equation tells you e.g. that in QED you keep on adding new
DISCUSSION 601
positrons with lower energy made through pairs because the right
hand side of this equation is positive.
In QED, is also increasing with t so that the rate at which
you produce pairs will be increasing. In QCD the rate of quark pair
production slows down but never stops. This is the difference
between QED and QCD. If you take a typical SLAC x - distribution
at a certain tl what happens is that in QCD you tend to accumulate
at low x at a decreasing rate while in QED you would do so at an
increasing rate. In both cases, if t2 > tl the structure functions
change as indicated in the figure 1. The results obtained from
these integral equations with the running coupling constant are
equivalent to the results obtained previously using the equations
for the moments of the distributions.
o i "
BUDNY:
Could you show how the quark distributions pO(x) and nO(x)
are de rived ?
CABIBBO:
We start with some ansatz for the wave function:
'f'abC 1 1
(x p~,
2 2
x P.a. '
33)
x PJ,
602 DISCUSSION
(a,b ,c are spin and internal symmetry indices). In order to use

this within SU(6) xO(3) we have to define a set of vectors where the
0(3) acts: e.g;
analogous
so that vI + v 2 + v3 = 0 and the vi's can be considered as relative

center of mass coordinates for the quarks. This has the further
advantage that
invariant under rotations invariant under Lorentz transformations,.

This gives you a relation between p~ distributions and x distribut-
ions.
2
In order to obtain a exp(-k.PL ) dependence we have to put
2
exp(-K~>. ). In fitting the x distribution there is the single
.

parameter
1
f3 = fA 2 .k which fixes the width of the x distribution, the
way in which the wave function drops as x '+ 1. With respect to spin
and internal symmetry we make an ansatz + b "':if. So there
are two parameters p and b. b is used to fit the ratio r:z.e.II(1-) . ~I
fieF'{l') ) lC
BUDNY:
Is it correct to view the three quarks represented by these

distributions as being large size clusters of quarks which could be
2
probed by low Q whereas the parton distributions p(x) and n(x) are
probed by high Q2 ?
CABIBBO:
Once we started we were thinking of some scaling distribution

of partons inside each quark. But now our point of view has devel-
oped and the parton distribution inside a quark should be considered
DISCUSSION 603
as a function of Q2 and computed from the set of integral equations.

2
Asymptotically free gauge theories fail you at low Q. SO any
picture has to be heuristic. We make a fit to this picture at a
certain Q2 and then calculate what happens for higher Q2. To lower
2
Q we cannot extrapolate because in this calculation our functions
will behave in an uncontrollable way. In order to do it in a
reliable way you should know your structure functions at high Q2
with infinite precision which you don't.
UEMATSU:
Asymptotically free gauge theories give us a field theoretical

explanation for scaling violation. But another possible explanation
is the case of a non-trivial fixed point where scaling is violated
powerlike in Q2. The present experimental data is not accurate
enough to distinguish between them. Could you make some comment on
this point ?
CABIBBO:
2 a
In the presence of a fixed point you have a (Q) - behaviour
where a might be a small number. So from experiment you cannot
distinguish between this and 10g(Q2) - behaviour. You can only put
limits on the power a. But our point of view is really different:
We believe in QCD and compute what happens; if we find it agrees
with experiments, we are happy, if not, we have to do something else.
UEMATSU:
2
If the parton distribution function depends on Q then the
2
fragmentation function is also expected to have Q - dependence.
What is your opinion on this point ?
CABIBBO:
I do not want to speak on fragmentation functions. But anyway

scaling and the use of the parton model in hadron physics is an
604 DISCUSSION
approximation itself. We don't even know how good an approximation

it is. So I don't think that it is really worthwhile to go to the
very fine details, since the scaling violation effects are anyway
small and the reliability of the approximation is uncertain.
TAO:
I understand that the Altarelli - Parisi analysis of QCD is

based on assumptions on quark distributions. Many doubts can be
raised on their validity but the fact that the results you get are
exactly the same as those of Georgi and Politzer using more compli-
cated mathematical methods appears to justify your analysis. I have
always been surprised that a very naive simple parton picture has
always allowed to calculate results equal to the ones from compli-
cated computations. For example, scaling has a very simple inter-
pretation in terms of parton language. Is it because current algebra
on the light cone completely justifies a parton picture ?
CABIBBO:
The justification comes from the results you get. They are the
same as in the usual analysis. For example in QED once you have the
Weizsacker Williams approximation it is a good thing because you can
compare it with exact calculations. If you ask a mathematical
question you cannot ask for an intuitive argument. I gave some
intuitive argument for these equations. Incidentally those can be
derived from Georgi's and Politzer's equations directly, because if
you solve the equations for the moments and then make an inverse
Mellin transformation, you will arrive at those equations after a
complicated and mathematically more convincing tour. This is act-
ually the way these equations for quark distributions were origin-
ally found by Parisi.
DISCUSSION 605
PREPARATA:
One must be very careful in applying this Altarelli-Parisi

picture to the actual physical world. As far as we know its appli-
+ -
cation can be justified only for e e annihilation, and deep inelas-
tic physics, where the Light Cone expansion can be used to inter-
polate from the physical region to the deep Euclidian region, where
the renormalization group gets stuck.
Beyond this, and in particular in the lepton pair production
case, it is totally unjustified, and I don't see what can be learned
from either its failure or its success.
HINCHLIFFE:
The Altarelli-Parisi calculation of the scaling violation can-

not predict by itself. The running coupling constant must be inserted
in the equations. Only when the parton ideas are supported by
Renormalization group analysis can they be believed. In particular
parton model analysis would imply scaling violation in the Drell-Yan
process and in parton fragmentation functions. These are not supp-
orted by asymptotic freedom.
But given the experimental fact of scaling violation in deep
inelastic muon scattering, the Drell-Yan formula loses predictive
2
power as the parton distributions inp.-p scattering are Q dependent.
What Q2 do you use in the Drell-Yan formula ?
The renormalization group analysis tells you nothing, and to
2
put Q =M or s is a hypothesis supported only by the parton
model
J"r
arguments.
CABIBBO:
This is a big problem which Q2 you should use, whether the mass
of the lepton pair or s or anything between. The Drell-Yan process
is the less well justified example among parton model applications
606 DISCUSSION
because it cannot lean on arguments based on the Light cone

expansion.
CARTER:
How did you get a pion wave function from that of a proton ?
o 0
Are u (x) and p (x) the same in both cases? What tests do you
have of this wave function ?
CABIBBO:
o 0
u and p will certainly be different. For the wave functions,
I best address you to the papers quoted in the lecture. Given a
wave function, I wish to recall how 1t leads to a K
'Ie
-f mass
difference formula:
~ 2. (~~ - ~~) < ~ >tt-t.r

.At _
JVI~ - (~ ~ - ~:) < >~esO'" *
(*>\'. t.'t0~
So, very roughly:
, (M~ -M~) : (M:_-M~)

z:
l. .... J ,--"",-.- ~
,2. ,
According to the rough evaluation indicated above, ~= 3/2, which

is in the right direction. If you take the wave function the
factor ~ increases to 1.9 which is nearly perfect. However any
distribution which is not peaked at x=l/2 or 1/3 would push the
factor 3/2 upwards. Therefore this is not a critical test of our
wave function but only a test of this description of mass
differences.
TOTAL CROSS-SECTIONS OF NEUTRINOS AND ANTINEUTRINOS
IN BEBC IN THE ENERGY RANGE 20-200 GeV
*)
P. Renton
Department of Nuclear Physics
Oxford, England
1. INTRODUCTION
This paper describes preliminary results obtained from an ex-

posure of the 3.7 m diameter bubble chamber, BEBC, filled with a 74%
molar NeH 2 mixture, to the CERN narrow-band dichromatic, V and V
beams. The results presented are the work of many people in a col-
laboration involving laboratories in Aachen, Bonn, CERN, Imperial
College Lond, Oxford and Saclay. Many of the results presented are
also given in Ref. (1), which gives a list of the authors.
The main aims of the experiment were to study:
a) crVN and crVN as a function of neutrino energy in the range 20-

200 GeV;
b) the differential cross-sections d 2 cr/dx dy, where x and yare
the Bjorken scaling variables, to extract the structure func-
tions and to test charge symmetry;
c) neutral currents;
d) dilepton production, ~~, ~e and any associated strange particles.
*) Aachen-Bonn-CERN-London-Oxford Sac lay Collaboration.
607
608 P.RENTON
This paper deals mainly with the determination of the total

cross-sections and the results derived from them. The essential
ingredients in the cross-section determination are described in the
following sections and are:
a) The number of neutrinos at BEBC and their distribution in ener-

gy and radial distance from the beam axis. This requires an
understanding of the beam (Section 2) and the flux monitoring
system (Section 3).
b) The number of neutrino interactions in BEBC and their separa-

tion into charged and neutral current events. The separation
is achieved by using the external muon identifier (Section 4).
c) Separation of charged current events into those produced by

neutrinos from pion decay and those from kaon decay (VK).
(V~)
This requires detailed measurements of the events (Section 5)

and permits an exploitation of the properties of the narrow-
band beam to cover the entire energy range of the neutrinos
(20-300 GeV).
2. NARROW-BAND BEAM
The following discussion is intended merely to illustrate the

principle of the narrow-band beam and the various technical reports
2)
should be consulted for details Figure 1 shows a schematic lay-
out of the main components of the beam. Protons of momentum 400
GeV/c are ejected from the CERN SPS (23 ~sec spill) and irradiate
the neutrino beam target. A dipole plus quadrupole beam channel is
used to select secondary particles of the required sign and momentum.
The momentum selected by the secondary beam was 200 GeV/c (5%) and
the beam had a divergence of about 0.2 mrad. The secondary particles
TOTAL CROSS SECTIONS OF NEUTRINOS AND ANTINEUTRINOS
609
125m 300m 184m

. 1.
I
1
torget I
1
.. .!!'~-o--
1
-----
400 GeVlc
protons
t t
Oecoy Steel shield InO with
1
BEBe
tu nnel 00 ps conlo l nlno
solid state detectors
Fig. 1 Schematic layout of CERN narrow band beam

610 P. RENTON
are directed into a decay tunnel, where the TI and K mesons in the
beam decay to produce neutrinos. About 3% of the TI mesons and 18%
of the K mesons decay in the 300 m long tunnel and the main decays
producing neutrinos are
+
-rr- -+ f.-+
In addition there is a small background from three-body lep-

tonic decays (K + ~v and K + TIeV ) which can easily be calculated.
~ e
There is a further source of neutrinos reaching the detector which
arises from decays of TI and K mesons before they are sign and mo-
mentum selected. This so-called wide-band background is mainly low
energy. Hence the beam is predominantly v~ with small V~ and Ve
components.
The kinematics of the two-body leptonic decays of TI or K mesons

is shown in Fig. 2 and leads to a
unique relation between the energy
R of the neutrino (E) and the decay
angle 8. If EM' PM and ~ are the
energy, momentum and mass of the
parent meson and m is the muon
~
Fig. 2 Kinematics of TI or
~2 mass, then:
or K decay
~2
E =
:J. [EM - PM cos ~J
~ PM (!- mt~/tm~)
1+ (PM/IMM)2f12
TOTAL CROSS SECTIONS OF NEUTRINOS AND ANTINEUTRINOS 611
where y = PM/~'M
and E
max
is the maximum neutrino energy (8 0).
For kaon neutrinos EK
max
= 0.95 Pm, whereas for pions, ETImax
If the finite length of the decay tunnel, together with the

divergence and momentum spread of the beam is taken into account,
then the profile of the beam at BEBC can be calculated and is shown
in Fig. 3, where R is the radius perpendicular to the beam axis.
3. DETERMINATION OF THE NEUTRINO FLUX
The decay tunnel is followed by shielding which consists of

184 m of steel followed by 170 m of rock. The shielding serves to
absorb the muons associated with the neutrinos and it also contains
gaps, at various distances into the stp-el. These gaps are equipped
with solid state detectors which are used to monitor the muons in
the shielding in order to determine the neutrino flux 3 ) . The
ZlK Mean energy of events

150 \ 90% of all events at a given
\ \ radius fall between !he
broken lines.
Rem
Zl7r \ \
\ \
100 \ "- \ \
\ \ "- \
\ "- '\. "- \
\ '\.. "- \
50 \ '\ "- \
\ \ "- \
\ \ \
\ \ \
\
I
a 150 250
E, GeV
Fig. 3 Beam profi le at BEBC

612 P. RENTON
detectors are distributed in a radial fashion around the beam axis

and in addition there are movable detectors which are used for cali-
bration.
The distribution of the intensity of muons as a function of

radius and distance into the shielding depends on~
a) the intensity of the parent beam;

b) the divergence and momentum spread of the parent beam;
c) the relative number of muons from TI and K decays;
d) the range-energy relation of muons and the effect of multiple
Coulomb scattering in the shielding;
e) any other sources of muons.
The radial dependence of muons from K and TI decay is such that,

in the central region, the distribution is relatively insensitive
to the number of kaons. The radial shape is, however, sensitive to
the beam divergence and Fig. 4 shows the expected radial distribu-
tion for different beam divergences together with the measured
points for gap 2, which is at 30 m depth. The K/TI ratio in the
parent beam has been determined independently4) to be K+/TI+ = 0.16
0.015, K /TI- = 0.054 0.003.
Using the beam parameters together with the range-energy rela-

tion and the mUltiple Coulomb scattering of muons in steel, the dis-
tribution of muons in the various gaps can be computed and is shown
in Fig. 5, together with the measured points. The curves are nor-
malized at one point only (central point in gap 2, 30 m depth) and
good agreement is obtained. The flattening of the distributions as
the depth into the shielding increases is due to mUltiple Coulomb
scattering.
A correction of (-7 2%) to the muon flux was made for "trapped"
muons from pion and kaon decays in the beam line (whose neutrinos miss
BEBC) on the basis of runs with the collimators of the momentum slit
closed.
..
Q
N
~ 2
......
5 10 15 20
RADIUS em
Fig. 4. Radial dependence of muon flux, compared with predictions

as function of beam divergence.
614 P. RENTON
/
.- 30m
""---
"'-
.....
2
1

-0
..~ \ --
10 40 20 o 20 40cm
Fig. 5. Muon horizontal distributions as function of depth.
The neutrino flux at BEBC is then calculated from a knowledge

of the number of particles in the shielding (which is continuously
monitored), the Kin ratio and the parameters of the beam.
An absolute calibration of the solid state detectors has been

made using emulsion stacks. Figures 6a and 6b show the results of
a scan made for tracks in the emulsions exposed in gaps 2 and 3.
In addition to the obvious peak due to muons in the forward direc-
tion there are backgrounds due to cosmic rays {determined in sepa-
-t
0
-t
}>
r
(')
:0
IOOO .Jr\ ANGULAR DISTRIBUTION IN --j 0
en
VERTICA L 250 fm EMULSION. en
en
CENTRE CAP (50m Il::PTH ) m
(')
-t
Muons (5
zen
0) b)
100 0
w "Z
...J m
C> C
d EJectrons ( Messel 6 Z -t

10
a: 0:: Z
...
0- W 0
a.. }>
..... ...... z
'"u" en ...... ..... -t
4 1
a: ::.:: ,, / ............ Z
to- \ u , ..... , m
... <I
\ '" 0:: E l eclrons C
...--t- - ...l- f ....-.... .
t- -t
:0
z
\\1r 'f, 0
I "-
en
I
I
1 0 1
50 100 200 300 400
ANGLE ""od ANGLE mrod
Fig . 6
0-
01
616 P. RENTON
rate exposures) and from 0 rays produced in the shielding. The

curve shown uses the results of a calculation by Messel and
Crawford 5 ) and shows the distribution is well understood.
The systematic error on the neutrino flux is estimated to be

~ 7%, the contribution from the absolute calibration being 4%.
4. BEBC AND THE EXTERNAL MUON IDENTIFIER
A diagram of BEBC and the External Muon Identfier (EMI) is

shown in Fig. 7. The bubble chamber has a superconducting magnet
giving a magnet field of 3.5 T and is filled with a 74% molar Ne/H2
-3
mixture, which has a density of 0.71 g cm and a radiation length
of 42 cm, to be compared to the chamber diameter of 3.7 m. The
combination of a large magnetic field and long track lengths leads
to good momentum resolution for charged tracks, this being ~p/p = 4%
for p = 100 GeV/c and a 2 m track length. The fiducial volume used
was 17 m3 (mass 12 t) and was chosen so that the minimum downstream
path length from an interaction was 0.5 m.
The EMI 6cons~sts

). of a . 1
s~ng 1
e pane 0f 49 mu 1"
t~w~re propor-
tional chambers, each 3 x 1 m2 area and mounted about 7 m from the

centre of BEBC. The angular range covered by the EMI was 80 0 in
the horizontal plane and 60 0 in the vertical one. The wire cham-
bers are separated from the chamber liquid by 0.5 - 1.5 m Fe absor-
ber. From the measurement of a track in the bubble chamber an ex-
trapolation is made to predict where the particle should hit the
EMI. This prediction has an uncertainty due to the measurement er-
rors and also due to multiple Coulomb scattering in the absorber,
depending on how much material is traversed. If a hit is recorded
in the wire-chambers, within the expected uncertainty of extrapola-
tion, then the track is identified as a muon. The probability of
hadrons traversing the shielding is negligible as is the background
from random hits for p > 5 GeV/c.
).l
- - --- ~ -. ---
E
"
.1<4.1. chambers
618 P. RENTON
The efficiency of the EMI for detecting muons greater than

5 GeV/c is 98% as determined from a study of about 2000 muons of
high momentum which traversed BEBC.
5. SELECTION OF THE EVENT SAMPLE
The film was double scanned (efficiency ~ 99%) and the events
were measured using film plane digitizers and on-line geometrical
reconstruction. Figure 8 shows an example of an event. For each
event the following were measured:
a) the tracks of all muon candidates (non-interacting tracks leav-

ing the chamber);
b) the tracks of charged hadrons (including the secondaries of in-
teractions if these were necessary to determine the momentum
of the track);
c) associated Va's and neutron stars, these are retained after
measurement if unambiguously associated with the neutrino in-
teraction;
d) any primary electron tracks;
e) any primary y rays (pairs or Compton electrons), plus brems-
strahlung y rays if they were necessary to determine the y ray
momentum more precisely.
Some of the events were confused, generally because one or
more charged tracks interacted close to the vertex, and for these
events as many tracks as possible were measured including all leav-
ing tracks.
The events were classified as charged or neutral current events

as follows:
a) charged current events - at least one leaving track greater

than 5 GeV/c identified as a muon by the EMI;
I
I
\(/ .
,I
I
\. ~ "
"0 \ .
I
\ i
\ / .; . ".
1',! :\ ~ ~",
!, ~
\ . . ---""'-
" ~. . .. .
.'
\\ .,
. '. \.
Fig. 8 Vertex region of an event in BEBC
b) neutral current events - the remainder, i.e., events with no

leaving tracks or with leaving tracks either less than 5 GeV/c
or leaving tracks greater than 5 GeV/c but not identified as
muons by the EMI.
Figure 9 shows a scatter plot of the visible energy of charged

current events versus the radius from the beam axis. Also shown are
the contours of the regions inside which 90% of Vn and VK events
should lie. On account of measurement errors and loss of some neu-
tral hadron energy, the observed energy distributions are broader
than those expected for perfect resolution. The magnitude of the
loss of hadronic energy was estimated in several ways:
620 P. RENTON
NEUTRINO
150 150 iiK \
E
u
.\ : \
\
lOO E
U 100 . \ .
rt: :.:
rt: .,
50 50 \ .,. :\
0
200 0 50 00 150 200
E GeV E GeV
Fig. 9
a) by studying the transverse momentum of the hadrons. which

should balance that of the muon in charged current events;
b) by studying the relation between the measured hadronic energy
EH and that calculated from the beam parameters, namels EHB =
= E\! - E11 , where E\! is the mean (TI/K) neutrino energy, for the
radius of the event, and E is the muon energy (well determined);
11
c) from measurements of film of 70 GeV/c TI- interactions in the
same Ne/H 2 mixture obtained in special runs.
These methods are all compatible and indicate a mean loss of

hadronic energy of about 20%, mostly due to unseen neutral particles.
Taking this effect into account it is possible to make a clean sepa-
ration of the events into those coming from TI and K neutrinos, with
errors that are smaller than the statistical errors in the sample.
In order to calculate the cross-section as a function of ener-

gy, the corrected visible energy was used for the \! events, whereas
TI
for the \!K events the mean kaon neutrino energy computed from the
beam parameters and the position of the event in the chamber was
used.
6. THEORETICAL FRAMEWORK FOR THE DESCRIPTION

OF CHARGED CURRENT INTERACTIONS
The inclusive charged current neutrino and antineutrino inter-

actions on nucleons, namely the processes
v,. + N ... r +
~+N +f/ ... .... .

can be des cribed by three variab les, name ly,
E = the incident neutrino energy
(p - P )2 = the square of the four-momentum transferred

V 1..1
between the leptons
~ 2EE1..I (1 - cos 81..1)' where 81..1 is the muon scattering angle

and
V =E - E1..I = the energy transferred between the leptons.
Alternatively the variables can be taken to be
E
2
x= q 102M'"
~= -E v JL
E
where M is the nucleon mass and x and yare the (dimensionless)
scaling variables.
For an isoscalar target, in the approximation that the Cabibbo

angle 8 ~ 0, and assuming scaling, then the differential cross-
c
section can be written in terms of structure functions in the
Bjorken formula 7 )*)
*) For a more detailed account of the theory of neutrino nucleon

scattering see. for example, Ref.(8).
622 P. RENTON
This can be simplified by use of the Callan-Gross relation 9 ) 2xFl

= F2, expected for spin ~ constituents of the nucleon
(2)
Defining B fXF3 dx/fF2 dx, this gives the differential cross-

section as a function of y
The charged-current scattering process has also been interpre-

ted in terms of the quark parton model. In this model the nucleon
consists of three "valence" quarks plus a "sea" of quark-antiquark
pairs (u~ + dd + s~ + cc + bb + ) plus gluons. This "sea" is
expected preferentially at small x values.
In terms of the quark parton model with spin 1 constituents,

the differential cross-section can be written
for \) (4)
for \i (5)
where Q(x) [Q(x)] are the quark (antiquark) probability distribu-

tions. These formulae contain the results coming from angular mo-
mentum considerations that neutrino-quark or antineutrino-antiquark
scattering gives a distribution which is uniform in y, whereas
neutrino-antiquark or antineutrino-quark scattering gives a (l_y)2
distribution for V-A coupling.
Comparing the above formulae, one can express B in terms of

q(x) and q(x), the fractional momentum of the nucleon carried by
quarks and antiquarks, respectively,
q(x) 'teX)
SOc) :: (6)
q (X) ..,. ~ ( )()
The ratio of the antineutrino to neutrino cross-sections on

nucleons can be obtained by integrating the relevant formula (3, 4
and 5) giving
G"v 1- B
R= - (7)
~ 2. + 8
If scaling was exact and the simple parton picture correct,

then the same values of B would be obtained from the neutrino and
antineutrino y distributions and from the cross-section ratio R.
Further, this value of B should not be a function of neutrino ener-
gy. However, the value of B and hence the shape of the y distribu-
tions can vary with x. Indeed, one would expect, in the quark par-
ton picture, a vlaue of B near unity at large x, where valence
quarks should predominate, but a lower value of B at small x where
the effect of the "sea" is greater.
624 P. RENTON
Deviations from scaling could arise for various reasons. For

example:
a) Intermediate vector boson W. This would contribute a propa-

gator term 1/(1 + q2/M2)2 and would lead to a decrease in o/E with
w
increasing E. The effect would be more pronounced in neutrino than
in antineutrino because the mean value of q2 is higher for the for-
mer.
b) New flavour production. In addition to charm there could ex-

ist other quarks which would affect both the neutrino and anti-
neutrino cross-sections. For example, a b-quark with right-hand
coupling has been invoked to explain anomalous effects observed in
V interactions. Such a quark would give a component which is flat
in y for V compared to the "normal" (1_y)2, so the effect would be
large if it has equivalent coupling (i.e. not suppressed).
New flavour production produces an increase in o/E with in-

creasing E. If W is the mass of the hadronic system, then W2 - M2 =
= 2MEy (1 - x), so any new threshold will appear first at large y
and small x.
c) ASymptotically free gauge theories. From asymptotically free

gauge theories it is predicted that, as q2 increases, the distribu-
tion of Q(x) as a function of x will shrink towards the origin. The
effect on the V and V cross-sections has been calculated recently by
Hinchliffe and Llewellyn Smith 10). A parameter A (related to the
strong coupling constant) appears in the theory 'but for the values
of A considered, the authors conclude that 0V/E and 0V/E fall with
increasing energy, but the ratio o~/OV rises because 0v falls more
slowly. Also, they conclude that between 50-500 GeV. 0V/E and O~/E
cannot vary by more than 5% unless there are more than four quarks
or the W is unexpectedly light.
Table 1 -t
o
Neutrino and Antineutrino cross-sections i!r
C')
::n
11) o
Gargame11e BEBC (this experiment) ~
en
m
Energy Range Q
2-10 20-60 60-ll0 llO-150 150-120
(GeV) oz
en
V1f V1f VK VK o
aV/E 0.72 0.05 0.67 0.06 0.56 0.05 0.61 0.05 0.51 0.05
"Zm
C
136 123 161 97 -t
N - ::n
events
z
-
aV/E 0.29 0.92 0.26 0.03 0.25 0.03 0.32 0.04
oen

z
N - 90 87 73 o
events
z
- -t
Z
R = aV /av 0.40 0.02 0.39 0.05 0.45 0.06 0.56 0.07 m
C
-t
::n
JF2 (x)dx 0.48 0.04 0.45 0.04 0.39 0.04 0.43 0.04 z
oen
JXF3 (x)dx 0.41 0.06 0.39 0.08 0.30 0.08 0.24 0.08
B = JXF3JF2 0.86 0.04 0.86 0.10 0.77 0.11 0.56 0.12
Cross-sections (in units of 10- 38 cm 2 nuc1eon- 1 GeV- 1 ) are corrected to those for an isoca1ar
target. Errors quoted for the BEBC experiment are statistical. Systematic errors on all
cross-sections are estimated to be ~ 7%. Systematic errors drop out in the ratio R, except
for the v K point, where an uncertainty of 10% in the K+K- ratio should be included. ~
~
626 P. RENTON
7. RESULTS FOR TOTAL CROSS-SECTIONS

FOR CHARGED CURRENT EVENTS
Table 1 and Fig. 10 show the results for the neutrino and anti-
neutrino cross-sections. The low-energy Gargame11e data l l ) in freon
are also included for comparison 0

A ssum1ng avn/ avp = avp /0~n = 2,
small corrections have been applied to the Gargame11e results (3%)
and to the BEBC results (1%) so that all the data refer to an
isosca1ar target. The BEBC data incorporate other small correction
factors:
a) An upward correction of +2% to all points to allow for EMI in-

efficiency.
b) A subtraction for wide-band background (-3%) based on "wrong-

sign muon" events, to the cross-section below 60 GeV.
c) A subtraction for events due to neutrinos from K decay, of

).13
-2% and applied to the data between 60 and 130 GeV.
ot:,. GARGAMELLE
'" BEBC (this experiment)
S BEBC from muon spectrum
800.5
f Ii801.0
I
C\J
f I NEUTRINO
E
u 0.4
co
f")
'0 >----------.f..-------- 5 GeV/c, based on a flat y-

].1
distribution for neutrino events, and a (1 - y)2 distribution
for antineutrinos.
Figure 10 also shows the cross-sections evaluated from the

muon spectrum in the neutrino events. Any event containing a muon
of energy exceeding the maximum V energy for that radius from the
'IT
beam axis, is assumed to be due to a v K interaction. The number
of such events allows one to determine the average v K cross-sec-
tion, in terms of an assumed muon production spectrum, and, by dif-
ference, the average vrr cross-section. Two pairs of points are
shown for extreme assumptions on the muon spectrum characterized
by the values B = 1.0 and B 0.5, respectively. The cross-sec-
tions deduced from the muon spectrum confirm the values obtained
from the total energy distributions of the events.
It can be seen that 0V/E is fairly constant with energy, where-

as 0V/E falls significantly from the low-energy value.
Figure 11 and Table 1 show the data on the cross-section ratio

0_/0 as a function of energy. Although the ratio is consistent
V V
with a slow increase in energy, any such increase is smaller than
that reported by the HPWF group 12), which indicated a doubling of
the ratio by E = 60 GeV.
The cross-sections in the forward direction y ~ 0 are a test

of charge-symmetry and one would expect do /dy = do_/dy in this
V V
limit. The results obtained are, for y < 0.1 and in units of
10- 38 cm 2 /nucleon,
-LdG'v
E d~
=0.15! D.09 0.71 ! 0.10 .
Assuming B ~ 1, these values extrapolated to y o give
O. <3 ... 0.2.

628 P. RENTON
t'
I
0.6 t- R =o-il/o- 1/ -
t
0.5 - -
0.4 -9 -
0.3- -
o GARGAMELLE
0.2 f- BEBe -
0.1 ...... -
I I I
50 100 150
1.0 JxF3dx 2 (0--0'")

B=--=
0.8 <?
f
!F2dx (0-+ iT)
0.6
0.4 t
0.2
o
E GeV
Fig. 11
Furthermore, from the average of the forward cross-sections, one

obtains
(dC'~!5 of d~/ci ~)
O. 79 :t O. f) 7 .
.tE.
This value corresponds to fF 2dx = 0.51 0.05 in Eq. (1). It is
determined from events of y ~ 0.05, and thus of q2 ~ 2 GeV 2 . It
can therefore be compared with the low q2 data from Gargamelle 11 )
1 3)
yielding 0.48 0.04; and with the SLAC e-d data via the quark
ed
model; (9/5)fF dx = 0.52. In summary therefore, the results
are consistent with both charge symmetry and the absolute values
of the neutrino and electroproduction cross-sections measured at
low q2.
In the case of exact Bjorken scaling, Eq. (1) would give

values for the structure function integrals fF 2 dx = 0.48 (0
V
+ 0 )/E
V
and fxF3dx = 0.96 (0V - 0 v )/E and their ratio B = 2(0V - 0 V)/
/(0 + 0_). These quantities are shown in Table 1 and Fig. 11 in

V V
order to facilitate comparison with the Gargamelle data as well as
B values previously obtained from y distributions under the scaling
14 15)
assumptlon' It is important to emphasize that if F 2 (x) and/or
xF 3 (x) depend on q2 = 2MExy, one does not obtain the factorization
of the x and y dependence of Eq. (1) and the values of the inte-
grals obtained from total cross-sections will be subject to small
systematic errors.
Figure 12 shows the results for the quantity <q2>/E as a func-

tion of E. Since
I I I I I I I
0,151-
w 0,10- 00 GARGAMELLE
-
ANTI NEUTRINO
"- X 15' FNAL
A
(\J SESG (this experiment)
c:r
V
I I I I I I I I I I I I
2 4 6 810 20 40 60 100
E GeV
Fig. 12
630 P. RENTON
the ratio q2/E = 2Mxy depends only on the muon track measurement,
and the total energy estimate enters only in the binning in E.
The BEBC data has been corrected for the p > 5 GeV/c cut, and pre-
~ 15)
vious results from Gargamelle and from the 15' FNAL chamber are
included. The lines indicate the empirical relation <q2>/E oc
O
oc E- .14 obta1ne
. d f rom an 1nt1re
. " 1y 1n"d epen dent ana1 0 f f1tS to
YS1S
" d " " b d 1 d . 15,16)
sca I 1ng eV1at10ns 0 serve 1n e ectron an muon scatter1ng
These results demonstrate again that there are significant devia-
tions from exact scaling q2>/E = constant) at neutrino energies
below 200 GeV.
Finally, the results on the charged-current neutrino and anti-

neutrino cross-sections from this experiment can be summarized as
follows:
a) The cross-section data in the region of low q2 are in agree-
ment with charge symmetry and with the previous measurements
on neutrino and electron scattering cross-sections in the
region q2 ~ 2 GeV 2
b) The total cross-section divided by energy is fairly constant

for antineutrinos and falls significantly for neutrinos.
c) Any increase of a-/a with E is due more to a decrease in

V V
a /E for neutrinos than to an increase of a_IE for antineu-
V V
trinos.
12)
d) The dramatic increase of a-/a reported by the HPWF group
V V
in the energy range covered by this experiment has been in-
terpreted in terms of a two-fold increase in a_ above the
V
threshold for production of hadronic states containing new
heavy quarks with predominantly RH coupling, and masses
~ 5 GeV. The results of this experiment are in strong con-
tradiction with this earlier work.
e) The dependence of av/E, av/E and <q2>/E on energy over the

range 2-200 GeV is consistent with the departures from
Bjorken scaling measured in electron and muon experiments.
-I
o
~
r
('")
::0
Table 2 oen
Dimuon events en
en
m
('")
-I
V
-
V
o
Z
en
- + Scatter plot of PfL- V PfL+ for 1/
CC EVENTS 517 lJ 251 lJ o"T1
dimuon events
Z
o=p..-p..+ m
- + C
lJ lJ = p..-p..+ KO -I
::0
120 x=p..-p..-
observed 7 1 Z
o o
u en
- background - 2.0 - 0.25 ......
>Q)
z
------------------- ------------- ------------ t!>
o
80 PfL-= PfL+
Net signal 5.0 3.0 0.75 1.1 z
':I.. -I
0- Z
lJlJ/lJ rate (%) 1.0 0.6 0.3 0.5 o ./ m
./ C
./ -I
associated VO 4 KO - 40

x
/
./
./
./

/' Z
::0
// o
en
+ + - f- ./
lJ lJ (V) lJ lJ (V) o ././
~
./
./
observed 1 0
o
. 20 40 60
- background - 1.2 - .3
PfL+ GeV/c
------------------- >-------------- ------------
Net signal o 1.2 0 0.3
Fig. 13
0-
w
632 P. RENTON
Such effects have been interpreted in terms of asymptotically

. . 10,17) .
free gauge theor~es of strong ~nteract~ons wh~ch pre-
dict that, as q2 increases, progressively more of the frac-
tional momentum fxF3dx of the valence quarks will be trans-
ferred to gluons and the quark-antiquark sea. In this respect
the over-all decrease of 0 /E with energy for neutrinos is
V
especially significant, since any new threshold effects (c-
quarks, b-quarks, etc.) encountered at high energy should in
themselves increase this quantity.
8. DHmON EVENTS
In essentially the same data sample as used for the determina-

tion of the charged current cross-section, several examples of di-
muon events have been found. Both muons are required to be iden-
tified by the EMI and a minimum momentum cut of 4 GeV/c was ap-
plied to the muons. The preliminary results are summarized in
- +
Table 2. For neutrinos, seven ].1 ].1 and one ].1 ].1 candidates were
.. + - + +
found, whereas for ant~neutr~nos one ].1].1 and no ].1].1 events were
observed.
The ma~n source of background is from the decays in flight

of TI and K mesons, the muons from these decays reaching the EMI.
The backgrounds due to both hadron "punch-through" (absorber is
approximately ten interaction lengths) and random hits are small
and can be neglected. It can be seen that the ].1].1 event is com-
patible with being background, but there ~s a small net signal for
- +
].1].1 in V of 5 3 events, giving a rate of (1.0 0.6)%, with
respect to single ].1 events.
Figure 13 shows a scatter plot of p _ versus p + for the neu-

].1 ].1
trino dimuon candidates. On average, the ].1- is much faster than
+ - +
that of the].1 for the ].1].1 events p _/p +> ~ 5.5). This, to-
].1 ].1
gether with the observation of four KO mesons associated with the
].1-].1+ events and the ].1+Ko effective masses of 1.07, 1.43, 1.80 and
2.18 GeV, is consistent with the behaviour predicted for the decay
of charmed hadrons.
REFERENCES
1) Total cross-sections for charged-current neutrino and anti-

neutrino interactions in BEBC in the energy range 20-200
GeV. Aachen-Bonn-CERN-London-Saclay Collaboration.
P.C. Bosetti, H. Deden, M. Deutschmann, P. Fritze, H. GrassIer,
F.J. Hasert, J. Morfin, H. Seyfert, P. Schmitz, R. Schulte,
K. Schultze and H. Weerts (Aachen).
K. Bockmann, H. Emans, C. Geich-Gimbel, R. Hartmann, A. Keller,
T.P. Kokott, W. Meincke, B. Nellen and R. Pech (Bonn).
D.C. Cundy, J. Figiel, A. Grant, D. Haidt, P.O. Hulth,
D.J. Kocher, D.R.O. Morrison, E. Pagiola, L. Pape,
V. Peterson, Ch. Peyrou, P. Porth, P. Schmid, H. Wachsmuth
and K.L. Wernhard (CERN).
S. Banerjee, K.W.J. Barnham, R. Beuselinck, I. Butterworth,
E. Clayton, D.B. Miller, K.J. Powell and E.J.W. West
(Imperial College, London).
W.M. Allison, C.L. Davis, P. Grossmann, R. McGow, J.H. Mulvey,
G. Myatt, D.H. Perkins, R. Pons, D. Radojicic, P. Renton
and B. Saitta (Oxford).
M. Bloch, M. Debeer, W. Hart, J. Sacquin, B. Tallini and
D. Vignaud (Saclay). CERN/EP/PHYS 77-37, submitted to
Phys. Letters B.
2) J.V. Allaby et al., Proposals CERN/SPSC/T13-l7; SPC/345 (1973).
3) W. Venus and H. Wachsmuth, CERN/T-CL Int. 73-4 (1973).
4) J. May, G. Sigurdsson and G. Stefanini, Status report on

narrow-band neutrino beam running and monitoring, CERN/EP
Int. Report, 26 April 1977.
5) H. Messel and D.R. Crawford, Electron-photon shower distribu-

tion function, tables for lead, copper and air absorbers,
Oxford, Pergamon (1970).
6) C. Brand et al., Nuclear Instrum. Hethods l36 (1976) 485.
7) J.D. Bjorken and E.A. Paschos, Phys. Rev. 185 (1969) 1975.
8) D.H. Perkins, Rep. Progr. Phys. 40 (197}) 409.
9) C.G. Callan and D.J. Gross, Phys. Rev. Letters 22 (1969) 156.
10) I. Hinchliffe and C.H. Llewellyn Smith, Dep. Theoretical Phys.

Oxford, preprint 44/77 (1977).
11) H. Deden et al., Nuclear Phys. B85 (1975) 269.

T. Eichten et al., Phys. Letters 46B (1973) 281.
634 P. RENTON
12) A. Benvenuti et aI., Phys. Rev. Letters 37 (1976) 189.
13) E.M. Riordan et aI., SLAC-PUB-1634 (1975).
14) A. Benvenuti et aI., Phys. Rev. Letters 36 (1976) 1478.
15) J.P. Berge et a1., FERMlLAB-Pub-77/44-EXP (1977) and Phys.

Letters 36 (1976) 639.
16) D.H. Perkins, P. Schreiner and W.G. Scott, Phys. Letters

67B (1977) 346.
17) D.J. Gross and F. Wilczek, Phys. Rev. D8 (1973) 3633 and
D9 (1974) 980.
MEASUREMENT OF NEUTRAL CURRENT CROSS-SECTIONS
AND THEIR ENERGY- AND y-DEPENDENCE
H.P. Paar
CERN
Geneva, Switzerland
1. INTRODUCTION
I want to discuss the properties of semi1eptonic, weak neutral

current interactions. The reaction can best be summarized as
v
II
(vII ) + Fe ~ nothing + anything
Here, "nothing" may stand for V (v) but we do not know. The talk
is divided into the following parts: neutrino beam, detector, da-
ta reduction, total cross-sections, energy dependence and y-
dependence of the cross-section.
2. NEUTRINO BEAM
At the beginning of this year the CERN 400 GeV Super Proton
Synchrotron (SPS) and its associated narrow-band neutrino beam came
into operation. The properties of this beam are especially impor-
tant for the study of neutral current interactions because the to-
tal energy of the final state is unobservable, contrary to the si-
tuation in charged current interactions.
635
636 H. P. PAAR
The energy of the incident (V) neutrino can, however, be

V
lJ lJ
inferred from the properties of the beam and the radial position
of the interaction in the detector, as follows. The neutrinos are
produced by the decay of charge-selected pions and kaons in a paral-
lel beam of well-defined momentum, 200 9 (r.m.s.) GeV/c. One can
show for the K(n) + lJV decay that the neutrino energy E is uniform
V
between limits given by
(1)
with
= EB/~'
-2
1 - YB '
energy of the V in the parent's centre-of-mass (30 MeV
for pions, 236 MeV for kaons),
EB parent beam energy (200 GeV),
~ mass of parent (0.140 GeV/c 2 for pions, 0.494 GeV/c 2 for
kaons) .
Numerically, these limits are approximately
o~ Ev ~ 0.96 (0.43) EB for K (n) (2)
The neutrino energy and the decay angle in the lab. e are related
by
(3)
where e is determined from the ratio of R (the radial distance of

the interaction from the beam axis) and the distance of the decay
point from the detector. For a given e, Eq. (3) gives two solu-
tions depending upon E*, i.e., whether the neutrino came from a pion
or kaon decay. The finite transverse dimension of the detector li-
mits e to be at most ~ 3.1 mrad. Therefore, we do not observe the
MEASUREMENT OF NEUTRAL CURRENT CROSS SECTIONS 637
MONTE CARLO NEUTRINO EVENTS
8000 r - - - - - - - - - - ,
RADIAL DISTANCE
40-80 em
6000
4000
2000
o '-'--__--".-o!=--_----'''----'
8000r----------------,
RADIAL DISTANCE RADIAL DISTANCE
80-120 em 120 -160 em
6000
4000
250
- Ev (GE'V) - Ev (GE'V)
Fig. 1 Monte Carlo simulation of the incident neutrino

energy spectrum for neutrinos
full kaon-neutrino spectrum implied by Eq. (2). Figure 1 shows

the primary energy spectrum of neutrinos that enter the detector.
The resolution in E is limited by the resolution in the measure-
\)
ment of the radial distance R (a ~ 22 em), the uncertainty in the

position of the decay point inside the 300 m long decay tunnel
(a 87 m), the momentum spread (a = 9 GeV/c). and the divergence
(a 0.2 mrad) of the parent beam. This gives a total energy re-
solution of 15% for kaons and 40% for pion neutrinos.
638 H. P. PAAR
Ul
::I
'1-1<II"'
<II
Po
Po
<II
III
.r:
- E-<
'~rT"~"""fI!.
N
.
3. THE DETECTOR
The detector combines the function of target. hadron calori-

meter, and muon spectrometer integrally in 19 similar modules (see
Fig. 2).
Each module consists of multiple iron plates 3.75 m in dia-

meter. with a total thickness of 75 cm of iron and a weight of 65
tons per module. The hadron shower is measured by means of scin-
tillator sheets divided into eight horizontal strips inserted be-
tween the iron plates. The first seven modules are equipped with
scintillators placed every 5 cm of iron, the next eight samples
every 15 cm, and the last four every 75 cm.
The vertical position of the showers is determined by weight-

ing coordinates of the scintillator strips with their energy re-
sponse. The horizontal shower position is calculated from the ra-
tio of the pulse heights at the right and left scintillator edge,
making use of the light absorption in the scintillator sheets (ab-
sorption length ~ 2 m). The radial distance R of each event is
thus determined with a resolution of 0 = 22 cm.
4. DATA REDUCTION
4.1 Hadron energy cut
The trigger relevant to this analysis requires an energy de-

position of more than 3 GeV in the detector. Events with a hadron-
shower energy (E H) of at least 12 GeV are selected. This is well
above the hardware trigger threshold. It is chosen such that the
cosmic-ray background is reduced to a manageable level.
4.2 Event time cut and cosmic-ray background
For each event, the event time relative to the beginning of

the experiment gate is recorded.
640 H. P. PAAR
In Fig. 3 a histogram of the event time for antineutrinos

(worst case) is given. The protons are extracted from the SPS in
a single turn giving a 23 ~sec long beam spill. This is clearly
visible as the excess events between 20 and 40 ~sec. The flat
background is due to cosmic-ray triggers. The above hadron energy
cut reduces these, as can be seen in Fig. 3.
The contamination of neutrino events by cosmic rays can be

calculated from the events outside the beam spill. The difference
of the electronics dead-time for events inside and outside the
spill is taken into account. The contamination by cosmic rays is
significant only for hadron energies below 15 GeV and is determined
with good statistical and systematic accuracy. It is about 0.7%
for neutrinos and 5.4% for antineutrinos after cuts.
3500
3000
2500
2500
Jc:
~
w
1500
E HAo > 12 GeV

1000
500
o
Event-time (lLsec)
Fig. 3 Histogram of event time of antineutrino interactions

without and with a hadron energy cut EH > 12 GeV.
4.3 Fiducial cut
The vertex position along the beam direction is required to

be between 0.35 and 8.25 m of iron. The lower cut excludes pos-
sible neutron-induced showers in the front part of the spectrome-
ter, although we have no evidence for such an effect. The upper
cut allows for a sufficient penetration depth in the rear part of
the spectrometer to be used for the separation of neutral current
and charged current events. The over-all length of iron is
14.25 m.
To minimize problems of side leakage of hadron energy, the

radial distance R is required to be less than 1.60 m. This cut
seems rather generous (the apparatus has a radius of 1.875 m), but
it is not because the resolution in R is 22 cm.
4.4 Event length cut
For each event, an event length L is calculated from the scin-

tillator information. It is equal to the thickness of iron between
the vertex and the end of the event. measured parallel to the axis
of the detector. This in no way requires any reconstruction of a
muon that might be present, so it is not necessary to discuss the
reconstruction efficiency. In Fig. 4 the distribution of L is
shown for neutrinos and antineutrinos.
We see a strong signal of neutral current events with an

event length less than 2 m, corresponding to the length of a typi-
cal hadron shower. The region just above 2 m is rather flat and
consists of charged current events with a short muon track.
The background of charged current events in the neutral cur-

rent signal is seen to be small; thanks to the large diameter of
the detector, it is mostly due to stopping muons rather than to
muons escaping at the side.
642 H. P. PAAR
1600,r------=~---------------------------------__,
NEUTRINO EVENTS WITH EH> 12 GeV
Z1200
~
u
E
~
Cl:
w (a)
a. 800
(/)
.....
Z
~
W
LL
o
Cl:
~ 400
:::E
::l
Z I MONITOR REGION .. I
COSMIC MUON
/ BACKGROUND
120 240 360 480 600

LENGTH OF EVENT (em IRON)
600,---------------------------------------------.
ANTI NEUTRINO EVENTS WITH E H >12GeV
Z
o
'!
E
u
~
Cl:
w
a.
(/) 300
.~
w (b)
>
W
LL
o
Cl:
w
CD
:::E
::l
Z I MONITOR REGION .. I
240
LENGTH OF EVENT (em IRON)
Fig. 4 Event length distribution for

neutrinos (a) and antineutrinos (b)
MEASURMENT OF NEUTRAL CURRENT CROSS SECTIONS 643
A neutral current candidate is defined by the requirement

0.16 < L < L (metres of iron), where L is made dependent on
cut cut
the hadron energy. The parametrization is L
cut
= 0.75 + 0.35 In EH,
with EH in GeV. The lower cut at 0.16 excludes most of the cosmic
ray bacground (see Fig. 4). The neutral current losses due to the
lower and upper cuts in L are together of the order of 1%. L
cut
was chosen such that the fractional loss is independent of the ha-
dron energy.
4.5 Charged current background
The charged current events with an event length L < L must

cut
be subtracted from the neutral current candidates and added to the
charged current candidates. Their number is determined by a Monte
Carlo extrapolation from the measured number of events in the region
3.11 < L < 5.11 m. For this extrapolation one needs the event
length distribution of charged current events with a low-energy
muon, i.e., high hadron energy or y = EHIE v near 1. The y-distri-
bution of charged current events is for this purpose assumed to be
of the form 1 + a cc (1-y)2 for neutrino events and (1_y)2 + a CC for
antineutrino events with a CC = 0.1 independent of neutrino energy.
These distributions are an adequate description of the charged cur-
rent y-distributions near y = 1. Since we are interested in the
region y near 1, the above extrapolation is insensitive to the ac-
tual value of a CC ' Thus we have taken the same a CC for neutrino
and antineutrino distributions.
The small contribution of events where the interaction takes

place near the edge of the detector and the muon escapes, is taken
into account by a Monte Carlo simulation. The systematic uncertain-
ties in the extrapolation are estimated to be ~ 5% for neutrinos and
~ 8% for antineutrinos. Table 1 gives the results.
644 H. P. PAAR
Table 1
Data reduction for the ratio of neutral to charged current

inclusive cross-sections for EH > 12 GeV
Neutrinos Antineutrinos
NC candidates 10770 104 3314 58

Cosmic-ray background -59 7 -119 10
WBB background -286 126 -646 116
CC background -1493 64 -235 49
NC with L < 16 and L > L +150 +43
cut
K correction -1008 -154
e3
NC signal 8074 156 2203 l30
CC candidates 26509 163 6483 81

WBB background -239 117 -323 83
CC extrapolation +1467 64 +253 42
NC with L > L -134 -35
cut
CC signal 27603 211 6378 123
NclCC 0.293 0.006 0.346 0.021

(error only statistical)
NclCC (final result, 0.293 0.010 0.35 0.03
systematic error included)
4.6 Wide-band beam background
In the region between the proton target and the beginning of

the beam-forming elements, hadrons of both signs and with the full
energy spectrum are present. Their decays give rise to a flux of
both neutrinos and antineutrinos. A subtraction of the events ~n-
duced by the wide-band beam flux is necessary because of their

very different energy spectrum, peaking at low energies. It is de-
termined in special runs where the momentum-defining collimator,
located half-way downstream between the beam-forming elements, was
closed. Thus the hadrons which normally produce the neutrinos are
stopped there.
The main flux of the wide-band beam background does not point
to our detector; still, this sort of background constitutes a
major source of statistical and systematic error in our analysis.
A fraction of approximately 5% of the running time has been de-
voted to closed collimator studies. The events observed in these
runs are scaled by the ratio of the collimator open and collimator
closed fluxes and subtracted. The subtraction is of the order of
4% for neutrinos and 29% for antineutrinos (see Table 1).
4.7 !e3 background
Electron neutrinos produce charged current and neutral cur-

rent interactions in the detector. All of them simulate neutral
current interactions because the electromagnetic cascade generated
by the final-state electron is contained in the hadron shower. The
observed neutral current to charged current ratio is therefore af-
fected. The Ke3 decay is the only important source of electron
neutrinos.
This background is suppressed by the relative branching ratio

of the Ke3 and K~2 decays, and by the lower energy of the neutrinos
produced in three-body decays. The background is further reduced
646 H. P. PAAR
by the admixture of neutrinos arising from pion decay. Since the

relative amount of kaons and pions in the parent beam is measured
to be K+/n+ = 0.17 and K-/n- 0.05, the calculation of the Ke3
correction is straightforward. It does not involve any unknown
elements. Table 1 gives the Ke3 correction; it is 12.5% for neu-
trinos and 7.0% for antineutrinos.
5. TOTAL CROSS-SECTIONS
We are now ready to present the results. In Table 1 the data

reduction and the results on Rand R- are summarized. With the
V V
exception of the final result, the quoted errors are statistical
only. The quoted systematic error arises from the uncertainties
in the charged current background subtraction and the Ke3 correc-
tion.
The final result for the ratio of the inclusive neutral cur-
rent (NC) to charged current (CC) cross-sections on iron nuclei is
(NC/CC\ 0.29 0.01

EH > 12 GeV
(NC/CC)\) 0.35 0.03
These ratios are integrated over neutrino energy; the average inci-
dent neutrino energy is 110 GeV for neutrinos and 90 GeV for anti-
neutrinos. In order to make a comparison with other experiments
and theory, we need to undo the hadron energy cut. For this we
need the hadron energy dependence of R = NC/CC in a region where it
is not measured (E H < 12 GeV). I will discuss this later, but now
already state the result:
(R)cor 0.29 0.01

(corrected for
hadron energy cut)
(R_) 0.34 0.03
V cor
In Fig. 5 the comparison with other experiments is shown.

The agreement between all these is good! Also drawn is the predic-
tion of the Weinberg-Salam model (solid line).
Now that we have established contact with what was already

known, we turn to really new things.
,
12
I
0.8r
1.1
I
I
I
1.0
,I THIS EXP
o GARGAMELLE
I
0.9 Q7.j. b. HPWF
. I
CITF
I
0.8 I
I
1
0.7 +
Ry 1
\
0.6
0.5
0.4
0.3
0.2
01
0 0.1 0.2 03 04 0.5 06

Ry
Fig. 5 Comparison of Rv and Ru with the Weinberg-Salam

model. The dotted line assumes no antiquarks,
the solid line is for q/q = 0.1 and corrected
for the non-isoscalarity of the target.
648 H. P. PAAR
6. ENERGY- AND y-DEPENDENCE OF THE

NEUTRAL CURRENT CROSS-SECTION
6.1 Method
The final-state neutrino is not observed for neutral current

events, and therefore the total energy and the scaling variable x
are not measured (contrary to the situation for charged current
events).
The scaling variable y = EH/Ev involves knowledge of Ev ' which

we have, but as discussed in Section 2 there is the n/K ambiguity.
Therefore we proceed as follows.
We use the properties of the narrow-band beam to predict EH

distributions for neutrino and antineutrino interactions separate-
ly in bins of R. This is done for various elementary y-distribu-
tions such as constant, (l-y), (1-y)2. Then linear combinations of
these distributions are made and the coefficients are fitted to
have best agreement between observed and predicted hadron energy
distributions. The corrections to the experimental data are evalua-
ted and applied as discussed in Section 4, but bin by bin.
A direct determination of the neutral current y-distributions

~s difficult at this moment because it requires a thorough under-
standing of all resolution effects. We are working on this but it
is not ready for use. In the meantime we work with the ratio of
neutral current to charged current cross-sections for which these
uncertainties cancel to a good approximation because, as we shall
see, the neutral current and charged current distributions are very
similar. The distributions of the data after corrections are given
in Fig. 6 and the ratios R are given in Fig. 7. It can already be
concluded that the neutral and charged current distributions do
not differ greatly. The NC y-distribution appears to be steeper
than the CC y-distribution for v and flatter for V.
We now turn to the fits.
NEUTRINO EVEN TS ANT INEUTRINQ EVENTS s::
m
~
Cf)
C
::0
RADIUS 0- 1.0 em RA DIU S 1.0 - 80 em RADIUS O- I.Oem RADI US 1.0 - 80 em -' 600 m
s::
m
Z
-I
o"T1
00 Z
m
C
1000 -I
500 ::0
~
NC 200 r
('")
C
::0
::0
m
o Z
-I
4 00~~-----------------' ('")
1500 ::0
200 o
Cf)
RADIUS 80-120 em RA DIUS 80-120 em RADIUS 120-160 em Cf)
Cf)
300 m
('")
:::!
cc o
z
Cf)
100
1"//#"//
A ,... I 0
100 200
(a) - EM (G eV) - EH (GeV ) (b) - EH GeV - EH (G eV)
Fig . 6 Hadron- energy distributi ons f or corrected NC and CC events

0-
~
for neutrinos (a) and antineutrinos (b) -0
N C FOR NEUTRNOS ~ FOR ANTlNEUTRINOS 8;
CC o
o 5 t- RADIAL 04STANCE RADIAL 04STANCE 101- RAD4AL OlSTANCE RAD4AL DISTANCE

0-40 em 40-eo em 0 - 40 em 40 - 80 em
04 08
03tt .. .... ~ .. fi T ' ..!.. l"
02
01 l in2ew 'O 11. i ln,zSw ' 0 l'

015 0 15
V- A V- A
0
RA04AL 1+ RA04AL RA04AL DISTANCE

05 t- RA04AL OlSTANCE
80-120 em 04STANCE DISTANCE f 120- 160 em
80-120 em
120- 160 em I
04 08
03 1, l' t. 1I I j t .1 .. 0 6.
, T
,
.... _--- -
02 04h ~ "'~ I.YI
II 1 I I
01 02 'W
o 00 2000 100 200 o 100 200 0 100 200 ::c
(a) - EM (GeV) - EM (GeV) (b) - E. (G eV ) - E . (GeV)
~
Fig . 7 Ratio of NC to CC cross-sections as a function of the hadron energy for neutrinos (a) "'tI

and antineutrinos (b) . The solid line is the prediction of the Weinberg-Sal am model
:0
with sin 2 Sw = 0.24 . the dashed curve for sin 2 Sw = 0 . 35 . The dotted line represents
pure V-A .
6.2 E
-\)
dependence of Neice
We fit the ratio R = Neice separately for E > 100 and E <
v v
< 100. The results are given in Table 2. So we have no evidence
for an energy dependence of R. It seems that the neutral current
cross-section is a linear function of E (because the charged cur-
V
rent cross-section is), a property required by scaling. From now
on we assume for simplicity that R is strictly independent of ener-
gy, although a small energy-dependence of 10% (for v) to 20% (for
v) cannot be ruled out.
Table 2
EV dependence of Rv and Rv
RV R_
v
Ev > 100 0.29 0.02 0.35 0.06
Ev < 100 0.30 0.01 0.34 0.03
6.3 y-distribution
If the neutral current coupling consists of any combination

of S, V, T, A, or P, then its y-distribution is a polynomial in y
of at most second degree.
Anticipating the final fit result, we choose constant, (1_y)2,

and y2 terms as follows:
652 H. P. PAAR
gL + gR(1-y)2 + gsP y2
NC (v) (4)
CC
fL + fR (l-y) 2
gL(1-y)2 + gR + gsP y2
(5)
fL (l-y) 2 + fR
We have used fL 1.05, fR = 0.07, fL = 0.98, and fR 0.10.

These values are suggested by the quark-parton model as suitable
input parameters for describing the CC y-distribution, based on
the experimentally determined amount of antiquarks in the nucleon.
Many of the results are insensitive to the specific values of
these parameters.
We find
0.02 0.07 .
The fit does not determine gsP very well. A significant S or P

neutral current coupling would lead to a large y2 term, which we
conclude to be absent. Let us see what happens when we drop the
gsP term:
NC gL + gR(1-y)2
(v) (6)
CC
fL + f R(1-y)2
gL (1_y)2 + gR
(7)
f L (1-y)2 + fR
These forms imply that the NC interaction contains V, A terms only,

although a suitable combination of S, T, P can imitate it (confu-
sion theorem).
We fit Eqs. (6) and (7) to V and V separately for the results
of Section 6.4 and 6.5. To increase the accuracy of the fits, we
can make a common fit where gL and gR are required to be the same
in Eqs. (6) and (7). The latter is suggested by the quark-parton
model. We find
gL = 0.30 0.01
"common fit"
gR = 0.050 0.005
6.4 ~ and ~ without cut in hadron energy
From Eqs. (6) and (7) we find for Rv and Ro (integrate numera-
tor and denominator over y):
3gv
V
L + gR
Rv (V)
3fL + fR
v + 3gvR
gL
R-V
fL + 3f R
Using the g determined from separate fits to V and v, we get
RV 0.29 0.01
Rv 0.34 0.03
These are the numbers stated in Section 5.
6.5 Charge symmetry invariance
Charge symmetry invariance requires
at y = 0
It is known that this relation is satisfied by CC interactions to

10%.
654 H. P. PAAR
To check it for NC interactions, we let y + 0 in Eqs. (6) and

(7) and get
v + gRv
gL
(~~) y=o fL + fR
v
v
V + gR
gL
(~~)y=o fL + fR
v
Again using the g determined from separate fits to V and - we find

v,
(~~) y=o 0.31 0.03 (v)
(~~)y=o 0.30 0.03 (v)
Notice that Eqs. (6) and (7) appear to violate charge symmetry in-
variance for CC interactions:
This is caused by the fact that iron is not an isoscalar target.

So charge symmetry is observed by NC interactions as well to a pre-
cision of ~ 10%.
6.6 Si~nificance of the right-handed comEonent

of the NC interaction
If the NC interaction was purely V-A, the NC y-distribution for

contains a (1_y)2 term only because of the presence of antiquarks
in the nucleon. We expect in this case gR/gL '" f x(~ + d + s) dx /
/ f x (u + d) dx. Here ~ud = f x(u + d) dx / f x(u + d) dx and
x s dx / f x(u + d) dx are measured to be 0.07 and 0.03.
~
s = f 2
These values come from dimuon production and the ratio of V and v
CC cross-sections. So we expect for pure V-A:
If gR/~ is larger. it indicates the presence of a V+A interaction

wi th quarks.
Experimentally (Section 6.3)
This establishes the presence of a V+A term with a significance of

more than four standard deviations.
If we change the amount of antiquarks introduced in the fit in

the range 5-20%. the resulting gR/ gL changes wi th it in such a way
as to leave the significance of the results unaffected.
Figure 8 gives the fitted value for gL and. gR together with

the expectations for pure V-A. V or A, V+A. and the Weinberg-Salam
model.
0.30 r-..---,--,.,------,--,-----,----,.--,-----,
Ia = q /q =0.10 I
0.25
0.20
0.15
30- CONTOUR
(STATISTICAL ONLY)
I
0.10
V-A
0.10 0.15 0.20

" 0.25 0.30 0.35 0.40
9L
Fig. 8 Relation between left-handed and right-handed

NC coupling constants ~ and gR'
656 H. P. PAAR
6.7 Neutral current sea and the Weinberg-Salam model
So far we have used assumptions of a rather general nature

such as Lorentz invariance (to limit the interaction to S, V, T,
A, P) and scaling (so we can write down y-distributions as we did).
We can now specialize to a specific model: the Weinberg-Salam mo-
del. It contains only one free parameter 8vJ ' Through it the neu-
tral current coupling is specified as a certain combination of V
and A (vector and axial vector). Thus the y-distribution is speci-
fied:
( dO) '" a + b (1-y) 2 (V)
dy NC
fda) '" a(1-y)2 + b (v)

ldy NC
1 5
a =
'2 - sin 2 8
W
+ - sin 4 8
9 w
5
b sin 4 8
'9 w
Using these equations, we can calculate Rand R-. The result is

V V
the dotted line in Fig. 5.
There is one complication: the presence of antiquarks in the

nucleon. To take this into account we use the property that the
y-distribution is unchanged under the simultaneous replacements
(quark + antiquark, neutrino + antineutrino). The amount of anti-
quarks is taken as in Section 6.6. The result is indicated as the
solid line in Fig. 5 and in Fig. 8. The results of the fit are
given in Table 3. As a check, we have made a fit to the data with
sin 2 8w and the antiquark fraction as free parameters. We get 16
5% for the antiquark fraction. This indicates a significant con-
tribution of antiquarks to the NC cross-section.
Table 3
sin 2 8W for V, V separately and combined
V V V + V
sin 2 8w 0.24 0.01 0.22 0.03 0.241 0.007
If we vary the amount of antiquarks from 10 to 15%, then sin 2

8W changes by 0.01. Other systematic errors raise this to 0.02,
so we have finally
sin 2 8W = 0.24 0.02
This value is somewhat outside the world average of 0.33. It has

the interesting property that it makes the pure leptonic weak NC in-
teraction an almost purely axial current. The mass of the W boson
is 76 3 GeV/c 2
7. CHARM-CHANGING NEUTRAL CURRENTS
Because it was one of the "whys" of Professor Zichichi, I want

to say something about this. If a neutrino produces a charm-
changing neutral current interaction, we expect to see a ~ + that is
a "wrong-sign" muon in the final state. So we look for the reaction
+
V + Fe + ~ + X
We have plenty of these events, but we have also the wide-band beam
background (see Section 4.6). To eliminate it we make a cut:
Ev~s~ble
" > 100 GeV. There are no wrong-sign muons after this cut
out of 17,000 charged current (~-) events. There are 1100 neutral
current events with Ev~s~
, 'ble > 100 GeV. No event means less than
658 H. P. PAAR
2.3 events at the 90% confidence level. An upper limit of 2.3/

/(1100 x B x eff) ~ 0.03 results. Here B is the branching ratio
~ ~
of the charmed hadron into a muon. and eff is the efficiency for
detecting such a muon. We have used B = 0.15 and eff = 0.5 as in
~
the dimuon analysis.
So we conclude that the charm-changing neutral current cross-

section is less than 3% (90% confidence level) of the normal one.
8. CONCLUSIONS
1) The ratio of neutral current to charged current cross-sections

is 0.29 0.01 for V and 0.35 0.03 for v subject to the ha-
dron energy cut EH > 12 GeV.
2) The neutral current cross-section is proportional to E to 10%

v
accuracy.
3) The neutral current coupling is of the form (V, A), mainly V-A
(left-handed) but with a significant (four standard deviations)
V+A (right-handed) part.
4) Charge symmetry is observed for the neutral current interac-

tion.
5) The charm-changing neutral current cross-section is less than

3% of the normal one (90% confidence level).
We can add the following conclusions with respect to the Weinberg-

Salam model:
6) Agreement with the Weinberg-Salam model (both from Rv' Rv and

from the y-distributions): sin 2 8W = 0.24 0.02.
7) Indication of neutral current coupling with the antiquarks in

the sea, similar in magnitude to charged currents.
Acknowledgements
The work that is reported here is the result of the work of

the people in the CERN-Dortmund-Heidelberg-Saclay Collaboration*)
and many outside this group. I want to thank all of them for their
contribution.
*) Its members are M. Holder, J. Knobloch, J. May, H.P. Paar,

P. Palazzi, F. Ranjard, D. Schlatter, J. Steinberger, H. Suter,
H. Wahl, S. Whitaker, E.H.G. Williams (CERN); F. Eisele,
C. Geweniger, K. Kleinknecht, G. Spahn, H.J. willutzki (Uni-
versity of Dortmund); W. Dorth, F. Dydak, V. Hepp, K. Tittel,
J. Wotchack (University of Heidelberg); A. Berthelot, P. Bloch,
B. Devaux, M. Grimm, J. Maillard, B. Peyaud, J. Rander.
A. Savoy-Navarro, R. Turlay (Saclay) and F.L. Navarria (Bologna).
DIS C U S S ION
CHAIRMAN: H.P. Paar
Scientific Secretary: H. Reithler
DISCUSSION
CARTER:
I have a remark to add to this morning's discussion

of the ratio of Mz to Mw' In the standard Weinberg-Salam
model, the ratio of the overall magnitudes of charged
current (CC) and neutral current (NC) couplings is:
amplitude NC N
Q
~
1M.,
~ 'Z. '2. I"!.
c.~ 0 ..,
"= w
u'2.
I~
1-2./ M~
M
ampli tude CC --M'"":'1=-12.-c.o--:",'Z::":e::-"w-- == 0('.

The standard model gives r=l, but this is a consequence
of the assignment of the Higgs scalars to an SU(2) doublet.
In a general SU(2)XU(1) model,
where -1 ,
~= the m1x1ng angle (not necessarily equal to tan (~/~),
as in the standard model)
t~t+l)= the SU(2) Casimir eigenvalue of the Higgs multi-
y = the hypercharge of the Higgs scalars (plet
a = a numerical factor depending upon the definition of
the charge operator. In the standard model, ~ =2.
In this general model, the ratio of NC to CC amplitudes
is proportional to:
1'1'1
661
662 DISCUSSION
Now, experiment indicates that

amplitude NC = r . (group theOry) '-:--:03 0 5.
amplitude CC factors ~ ~u.;j ~
Since the factor in parenthesis (which is a function of

the SU(2) and U(l) quantum numbers and of sin 21) is just
group theory and necessarily of order unity. we surmise
that r is of order unity. Of course, it could be 2 or
5 or 1/2, but assume it is exactly unity. Then one can
show that the Higgs scalars must belong to a doublet
(standard model), a septet (t=3), or a representation
with t) 10. This statement is independent of the choice
of ~ (choice of charge operator).
I would like to add that there could of course be
more than one Higgs multiplet. But one should keep in
mind that if the absence of Higgs-mediated flavor-changing
neutral currents is to be natural, only one Higgs multi-
plet can be involved in the mass-generating mechanism.
Finally, it must be possible to construct a gauge-
invariant Yukawa coupling (V "'Rc:P~t.+'\IL.CP-YR. of the leptons to
the Higgs particles in order to generate lepton masses.
This means that the quantum numbers of the Higgs' must be
related to the quantum numbers of leptons. It is diffic-
ult to incorporate this constraint in the analysis. But
the conclusion is that, if you believe in the group
SU(2)XU(1), the neutral-current data may already be
telling us something about Higgs particles.
CABIBBO:
I could add a remark on this subject. If we assume
that the gauge group is SU(2)XU(1) and the usual assign-
ments for the quarks, i.e.: left handed particles in
doublets and right handed particles in singlets, the
neutral current jZ is determined in terms of aw , and
one is led to predictions for R~ and R~ of the form
R'I = f (~w). ( M.., )"""
~ ~C()\ew
R~ = ~ (Q..,) . (t-\.., \+
M2,cos9w )
fv and f~ are (under the above hypothesis) known function
DISCUSSION 663
of9wand of quark and antiquark fractions in the targets.

The standard Weinberg and Salam model predicts
Mw I (M z cos 6w ) = l,
so we can use the above predictions to have two independ-
ent determinations of &w. However (this was pointed out
by M. Veltman), it is possible to modify the model in
such a way that Mz~Mwl cos e...,.
A possible modification
is the addition of a triplet of Higgs particles to the
usual doublet.
It would therefore be interesting to analize the
da ta on R \1 and R" without assuming Mz =Mwl cos ew One
could solve the two equations above to determine both.sw
e
and the ratio Mz/Mw cos w ' to see with what precision the
prediction of the standard SW model is verified by data.
This analysis is more restricted than that indicated by
Carter, which would allow abandoning the doublet and
singlet assignments for quarks and leptons, as well as
the doublet assignment for Higgs bosons.
HINCHLIFFE:
The relationship between Wand Z masses, Mw/Mz=cos&w

is only valid if the underlying group is SU(2)XU(l)
(Weinberg-Salam). If the gauge group is extended to, say,
SU(2)XU(l)XU(l) there will be no difference in the charged
current, but now there are 3 neutral bosons (y, Zl, Z2)
and there are now 3 mixing angles (analogous to S"" in the
SU(2)XU(l) case). There will be a relationship between
Mzl, Mz2 and Mw. The 3 angles will be such, that
Mzl < Mw I cos~

Mz2 ') Mw I cos~
where~ is a mixing angle. This model has been proposed

by T.Weiler and G. Ross (unpublished) to explain the
absence of parity violation in the atomic physics experi-
ments. Another proposed model is SU(2)XSU(2)XU(l) :
again there are 2 Z's (Mohapatra and Sidhu; also Fritzsch
and Minkowski).
664 DISCUSSION
SANDA:
The modification of the relation Mw/Mz=cose w , due

to possible additional Higgs bosons has been considered
by B. Joglekar a few years ago, and it is published in
The Physical Review (1973,1974 ?).
LIPKIN:
Can the difference between the Harvard-Penn.-

Wisconsin-Fermi lab and CERN experiments be due to an
additional component in the beam, e.g. neutrinos from D or
F decays, neutral heavy leptons, or completely new objects:
PAAR:
Yes, such a possibility has been pointed out to us

by B. Richter. The HPWF detector is located at zero
degrees with respect to the proton beam, while our detectox
is off at an angle (in order to reduce wide band beam
background). If those or other objects are produced with
limited PT, they could miss our detector. This is not so
for HPWF. However, to explain the difference between the
experiments, one must assume an unreasonably high prod-
uction of these particles. Anyway, we will have an answer
to this very soon from our wide band beam run, for which
our detector is at zero degrees with respect to the
incident proton direction.
CHEN:
Regardless of theoretical models, from the energy

dependence of the charged current and neutral current
cross sections, you should be able to set a lower limit
on the mass of WX and Zoo Do you have any numbers?
PAAR:
Not at the moment, because our analysis of the

energy dependence of the individual ~ and ~ cross sections
is not finished.
CHEN:
Since the sampling of your detector is quite crude,

DISCUSSION 665
don't you have problems in measuring the energy of

electromagnetic particles such as \,~~ e etc.?
PAAR:
Missing energy does not cause a shift in hadron
energy, because the calibration constant makes up for it.
Of course, the resolution becomes worse. The calibration
constant is measured using a test calorimeter in a test
beam of pions and electrons with well defined momentum.
Electrons and pions give different calibration constants,
but the difference is less than 10%.
FERRERO:
10% of the muon tracks are lost by the reconstruction

program. I would like to know a) the reason of this
program's loss and b) the distribution of the lost events
( 5000) as a function of the radius or E~ , and the
influence of these missing events on charged current
analysis (no problem arises for the NCICC ratio, because
the muon track reconstruction is not required).
PAAR:
This loss is due to failures in pattern recognition
and trackfitting. We have measured by hand 1000 events,
taken from the 10% lost events. Their distributions look
the same as the ones for events that are reconstructed by
the computer. So the loss does not introduce a bias. I
might add that we have also checked a few hundred events
that were successfully reconstructed by the computer and
remeasured them by hand. No systematic difference was
found.
KNOBLOCH:
The reason for program losses are essentially two:

1. Muons which do not undergo a "smooth" energy loss in
the iron, but loose several GeV at one point, give rise
to problems in the fitting procedure. 2. Low curvature
(high momentum) muons which are produced near the down-
stream end of the fiducial volume traverse a few chambers
only, and thus make the fit difficult.
666 DISCUSSION
FISCHER:
On flavor changing neutral currents you gave an upper

limit of 2% with 95% c.l Have you taken DO_no mixing
into account? Would not this raise your upper bound?
PAAR:
No; it was not taken into account. If we knew
something about it, we could take it into account. It is
possible, that this could raise the bound.
TAO:
So far, nobody has measured the direction of the
hadronic shower. This is not important in charged current
physics, since the muon angle is (very well) measured, but
in neutral currents it is impossible to get the x distri-
bution. Measurements show that the resolution on the
angle of a hadron entering a calorimeter varies roughly
like l/logE or a+b/E. Our Monte Carlo calculations show
that if the angular resolution is 10 mr at EH=lOO GeV,
then it is possible to distinguish between (1-x)3, and
(1-x)2 or (1-x)4 distributions. But if it is 20 mr,
then (l-x) and (1-x)5 look very similar. To compute it,
you need to know the structure of the hadron shower;
whether the hadron energy is distributed among one, two,
or more particles; how the energy is distributed among
them. This knowledge is crucial for the measurement of x
distributions. Could you give me some information on this?
PAAR:
We have information only on the lateral spread of
hadron showers from our test calorimeter. A preprint
describing this has come out.
ZICHICHI:
K. Winter et ale have measured the direction of
hadron showers in a test calorimeter and it seems to
work.
DISCUSSION 667
RENTON:
My question concerns the charged-current events

which are hiding under the neutral current signal. For
what values of muon momenta does this occur and how
sensitive is the extrapolation you make to the B para-
meter of the CC events? Is this included in your error?
PAAR:
If you look at the event length plot, you see that

a typical hadron shower is about 2 m long in iron. A
muon looses about 1 GeV per meter, so the momentum of a
muon hiding in the hadron shower is :S 2 GeV / c
The estimate of the CC hiding under the NC peak needs

the y distribution near y=l. If there is a high-y anomaly,
we would be in trouble, but as you know from Knobloch's
talk, there is none. For example, the ~ CC Y distribution
is of the form (l-y) 2 + c(cc , which goes to c:A.u =constant
as y ~l. Similarly, for" we have 1+ olee (l-y) 2 -1 as y ~l.
So we use a flat y distribution for the extrapolation
from the region immediately to the right of the NC peak
in the event length plot to the NC region. The result
does not depend on B (or <lee ), so the error on B is not
included in the error of our result.
RESELE:
In your determination of the Weinberg angle and of

the sea contribution Nc(i.e. the fit to the y-distribution
under the assumption of a,b given by the Weinberg angle)
you got for \) induced NC 01.1"(= 0.35:t 0.14. Is not this
value disturbingly large? If you compare this value with
ot~c: = 0.17 0.02 of ;; induced NC, does not this put the
W-S model in a bad situation?
PAAR:
The error in the 0.35 is so big, that there is no

problem with this value. The reason the error is so big
is due to the fact that for v , olNCis determined mainly
from the (1-y)2 contribution to the cross section, while
for ~ it comes from the constant term in the y distribut-
ion. It is easier to determine a constant than something
668 DISCUSSION
mUltiplying (l-y)2 in a fit, so you get the errors as

they are. You can also see this effect in the fit of
a and b alone.
BUDNY:
I would like to comment on the result that wrong sign

muons are not seen with large Ehadr' Besides implying
that the WO cannot change a u quark to a c quark, it also
implies that there are no c quarks in the seas of nuclei,
since we expect diagonal WO coupling to c quarks. This
means that the production of J/~ in hadronic collisions
is not due to c and ~ quarks in the seas combining. Also,
it appears to contradict the prediction of asymptotic
freedom that co pairs in the sea become increasingly
important at high Q2.
PAAR:
From the 2% upper limit I gave for the ratio of charm

changing neutral current cross section to total neutral
f
current cfoSS section, you get that the ratio of 2 xc(x)
dx and x(u(x)+d(x dx (in Feynman's notation) is lesl
than 2% in the standard quark parton model.
The Q2 involved in this reaction is not so h~gh. It

is Q2 = 2MExy ~ 2xO.94xlOOxO.05xO.5 ~ 5 (GeV/c) , so it
says nothing about high Q2 behavior of ce.
SANDA:
Let us suppose that the narrow resonance at 9.5 GeV

found in P+P -+ trt-'- + X indicates the existence of a new
quark at mq N 5 GeV. Now, for example, if the new quark
couples to u and d quarks with the usual strength, we
already know from what you have shown us that the current
cannot be V+A. From theoretical studies of HPWF events,
5 GeV quarks may produce deviations from linear rise in
the total cross sections. These are two indications, that
your experiment may already say something about the inter-
actions of the "new quark". Do you have something to say
about this? If not, will you have it within a near
future?
DISCUSSION 669
PAAR:
The analysis of the energy dependence of the cross
sections is not finished, so we cannot say anything
about this at the moment. We will have this by the fall.
CHARGED V+A CURRENTS IN LEFT-RIGHT SYMMETRIC GAUGE MODELS*)
R. Budny
The Rockefeller University, New York,
N.Y. 10021, USA
Knowledge of weak interactions increased dramatically in 1956

and 1957. During those years parity non-conservation and the V-A
hypothesis for charged weak currents were proposed. In accordance
with the V-A hypothesis, the leptonic current appears to have the
form
La [V,t)'CI\. \"-Ys ,)1 -+ h,c, J (1)

.t: e) \4""
The nucleonic current is
(2)
for small momentum transfers. Due to the factor 1.25, this is not
exactly V-A, but the consensus is that the currents of the usual
quarks are V-A analogues of the leptonic currents:
*) Work supported in part by the U.S. Energy Research and De-

velopment Administration under grant contract number
EY-76-C-02-2232B. *000.
671
672 R. BUDNY
(3)
where de =d cos e + s sin e and se =s cos e - d sin e. The ob-

served current in Eq. (2) is thought to result from Eq. (3) via
dynamics.
I will argue, using a gauge model, that the V-A hypothesis

should be modified, that there can also be V+A currents which give
small contributions for small momentum transfers, but which contri-
bute equally with Eqs. (1) and (3) at large momentum transfers. In
this limit, the weak interactions become parity conserving.
Consider the standard Weinberg-Salam model l ) built on the gauge

group SU(2) x U(l). The left-handed fermions eeL = 1(1 - ys)e,
etc.] form doublets and the massive right-handed fermions [e R =
= ~(l + ys)e, etc.] are singlets under SU(2), i.e.,
( ~ }L ' ( ~ )L' (:~)L 1 >

e ~)I~)
u ~ fZ.)'"
(4)
(:9\' Ce\> C'b)L'J

This model contains QED and the V-A hypothesis which work very well.
It also predicts neutral currents which agree with neutrino scat-
,
ter1ng results 2) ,an d'1t can b e ma de compat1' be lW1t
' h CP V10
' 1 at10n
' 3)
There are two places where this standard model may be in

trouble. One is in bismuth and the other is in tri-muon production.
CHARGED V+A CURRENTS 673
The trouble in bismuth is that experiments measuring optical rota-

tions have found 4 ) that the parity violating effect is an order of
magnitude smaller than expected from the standard model. The mea-
surements suggest
(5)
where jA and J V are the axial and vector parts of the neutral cur-
rent and IBi> is the bismuth nucleus.
There are several ways to fix this disagreement. One ~s to

enlarge the particle multiplets. In vector-like models 5 ) , the elec-
tron occurs in left- and right-handed SU(2) doublets,
where N is a neutral heavy lepton. For such models, the neutral

current coupling to electrons has the form
(7)
so the axial piece cancels, and <eljA1e> vanishes to lowest order.
A problem arises when we include quarks in the vector-like

mode1 6 ). Then the model predicts equal deep inelastic neutral cur-
rent cross-sections for neutrinos and antineutrinos, but the CERN-
7)
SPS neutrino experiment teaches us
lD3b D.()~)
(8)
lD.ID!: 0.0\")
674 R. BUDNY
The only SU(2) x U(l) gauge models with conventionally charged

8)
quarks in doublets, lepton-quark symmetry, and naturalness are
the standard model and the vector-like model, so the results in
Eqs. (5) and (8) imply that we must give up at least one of these
conditions.
One possibility is to change the gauge group. Hany larger

groups have been considered. Examples are E7 9), SU(3) x U(l) 10)
11) )
SU(2) x U(l) x U(l) , SU(2)L x SU(2)R x U(l) 12-14 , and
15)
SU(4)L x SU(4)R x U(l) Hodels built on these groups have more
than one Wo , and have parameters such as mixing angles which can be
adjusted to fit the neutral current neutrino data, and to keep
<eijAie> close to zero. Also some of these models provide plausible
explanations of the tri-muon events.
Hodels built on gauge groups containing identical left and

right subgroups, such as SU(2)L x SU(2)R' can be termed ambidex-
trous, or racemic (which in chemistry means a mixture of an equal
amount of left- and right-isomers of a substance). These models
can be . ,
bu~lt e~ther w~th
.12,13)
, or ,
w~thout
14) 'f
man~ est
Ief t-r~g
' ht
symmetry. For models having this symmetry. the Lagrangian and

physical currents are invariant (except for finite radiative cor-
rections) under chirality reversal (Ys ~ -Ys). This transforma-
tion exchanges the V-A and V+A currents and exchanges the left- and
right-handed particles. Hodels with this feature have many attrac-
tive features, some of which are described below.
Consider the SU(2)L x SU(2)R x U(l) model with manifest left-

right symmetry. The fermions occur equally in left- and right-
handed doublets with the representations of SU(2)L x SU(2)R x U(l)
as follows:
(~ ,0)-1)
The electric charge is given by
(10)
The Higgs bosons can be assigned as follows:
(ll)
and
with the Higgs potential being left-right symmetric.
The symmetry can be broken spontaneously if the vacuum expec-

tation values are unsymmetric:
676 R. BUDNY
<X:> =- (:J <Af. = (:J

<SL> - 0 <f>It') =
G) (12)
and
+
The parameter b contributes only to the mass of W and can be R
chosen to make WR heavier than WL in order to make V-A dominate
over V+A. How much heavier WR must be will be shown below. The
parameters AL and AR contribute to the masses of the two ZO's.
With the (unnatural) choice AL = AR' the ZO's generate an effec-
tive neutral current interaction for electrons and nucleons which
is the sum of a jvJv term and a jAJA term. Each conserves parity,
so the parity violating effect in bismuth vanishes to lowest order.
In the rest of this talk, I will describe some results on the

V+A interactions in this model which I have obtained with Beg,
1 3)
Mohapatra and S1r11n The charged weak 1nteract10n results
from the fermion-W coupling of the form
(13)
where (V-A) is the sum of the currents in Eqs. (1) and (3). This
P
is manifestly left-right symmetric. If the WL and WR mix, we must
diagonalize the mass matrix:

WI) = 'W L CoS ~ - \N R 'Sly.,"t
W2 = WR<:.o5! + W L c;,-oI (14)
The effective interaction, for momentum transfers q2, can be writ-

ten
where
G = ~2.. r
(loS -$ i1"l -r') "Z. -t }z (c.o~! ",,"'ina -r Jz.
>li ~ T'Y112. - q '2. 8 m~-9'2.
(16)
and
1+tay..T
=
1 - tan!
For process involving neutrinos, this effective Lagrangian can be

written
678 R. BUDNY
J>~ft =G 1 {[ (V-A't;-~ (V+A)+] e)'P(1-'{~)y

~ (1+2t"+S) r i
-t
(17)
+ [S (YfA') \. '(" l"-A )''']
f r
e'ire.. \+)'5) V + h. c.}
with
i"; 1 - ttAA (18a)
1- 2\A-to '1A/>.
and
$; 1 + Zrt VA + 'lAA
1-2Yl VA ... '1 AA (18b)
This implies the following consequences for S decays and ~ decays:
i) !~!~~_E~!~~_!~E__~~~~Z_~~~_~_~~~~Z~
1 6)
The usual formulae for the rates continue to remain valid
if one makes the identifications:
(19a)
(19b)
( 19c)
e being the left and right Cabibbo angle.

=- 2yt"A / (1;- r\,,,, ') (20a)
(20b)
Here the suffixes F and G.T. mean transitions induced by the vec-
tor and axial-vector currents, respectively. (Pure Fermi and pure
Gamow-Teller in the allowed approximation). Also, the expressions
for the polarization are normalized in terms of the standard re-
16) .
su 1 ts der~ved ~n the V-A l~m~t.
17)
The standard parameters , p, 6, ~. and n. which determine
the spectrum are related to the interaction parameters as follows:
(2la)
(2lb)
s_ (2lc)
(2ld)
11,=0
(2le)
680 R. BUDNY
Table 1
Bounds on weak interaction parameters
Lower Upper
Quantity Experiment
bound bound
V-A
P(e)F/P(e)F 0.939 1 0.97 0.19 (a)
V-A
p(e)G.T./P(e)G.T. input 1 1.001 0.008 (b)
P(e) /P(e)V-A 0.959 1 1.00 0.13 (c)

).l ).l
P().l)iP().l)~-A 0.985 1 1.1 0.03 (d)
p input 0.750 0.752 0.003 (c)
0 0.750 0.750 0.755 0.009 (c)
~ 0.959 1 0.972 0.013 (c)
Tl 0 0 -0.12 0.21 (c)
m2/ml 2.76 00
tan Z;; -0.060 0.054
TlAA 0.813 1.23

-n VA 0.698 loll
r -0.052 0.052
s 0 0.135
a) See H. Frauenfelder and R. Steffen in "a, Sand y spectro-

scopy" (K. Siegbahn Ed., 1964).
b) J. Van Klinken, Nuclear Phys. 75, 145 (1966).
c) Review of Particle Properties, Rev. Mod. Phys. 48, No.2,

Part II (1976).
e) M. Bardon et al., Phys. Rev. Letters i, 23 (1961).
We can use Eqs. (2la) and (2lc) to express rand s defined in

Eqs. (18) in terms of p and~. The result is
1"= 3- 4 r (22a)
31i+4p
(22b)
Measurements of these decays constrain the parameters nVA and

nAA , or equivalently tan ~ and ml/mZ' Some measurements such as
Cabibbo universality require theoretical assumptions in their analy-
sis. First consider the unambiguous measurements. The best avail-
able experimental values are given in Table 1. Assuming a spread
of two standard deviations. these measurements constrain nVA and
nAA as shown in Fig. 1. The way the constraints work is as follows:
The average P(e)G.T. measurement in Table I, together with Eq. (20b) ,
implies
~&:r. ~&.'T.
< --;:::::===- (23)
1_ J1- ~&2..'T.
with PG T . 0.985. This gives the wedge in Fig. 1. Note the

V-A
theory gives an upper bound of 1 for P(e)G.T./P(e)G.T.' The e mea-
surement and Eq. (2la) constrain nVA and nAA to lie between the op-
posite branches of the hyperbola
(24)
682 R. BUDNY
0.985~ PoT~ I
2.0
1.5
0 .746 ~p!: 0.750
, ~~==========::~~~~~~~=========1-I .O ~AA
0 .994 ~ GI'GCO 9 !: 1.006
0.5
-2 .5 -2.0 - 15 -1.0 -0.5

~VA
Fig. 1 The values of nVA and nAA consistent with S decay

and ~ decay measurements are unambiguously con-
strained to the region abed, and can be further
constrained to a'b'c'd' using theory dependent
Cabibbo universality measurements.
With P.
m~n
= 0.746, Eq. (24) gives the boundaries ab and cd in
Fig. 1. Note p cannot be greater than 0.750. The measurement of
~ and Eq. (2lc) constrains nVA and nAA to lie within the tilted
ellipse which is too large to give anything new. There is also a
theoretical constraint:
(25)
which gives the boundary ad.

The allowed region, given these constraints, lies within the

Procrustean bed labelled abed in Fig. 1. The corresponding region
in the tan s- mI/m2 plane is shown in Fig. 2. For orientation,
the V-A point is on the boundary: nAA = -n VA = 1 and tan s = mI/m2
= O. These constraints give the lower and upper bounds on obser-
vables and parameters listed in Table I. Note that the mixing
angle s must be small, so the WI gauge field has predominantly V-A
coupling and the W2 gauge field predominantly V+A. It is remark-
able that the mass of W2 could be as low as 2.76 times the mass of
WI (at the point c). This means that V+A) interactions could ex-
ist at a relative level given by s in Eq. (22b) which could be as
high as 13.5%.
Ion ~
o b
005
,
0 b'
0 .1 0 .3 0.4 ~
0 m2
c'
d'
- 0 .05 d:r-----------------~ c
Fig. 2 The values of tan sand mI/m2 are constrained to

the regions abed and a'b'c'd' corresponding to
those in Fig. 1.
684 R. BUDNY
Experimental tests of Cabibbo universality provide theory de-

pendent and thus slightly ambiguous constraints on the parameters
V
by specifying G~ cos 8/G S' Using measurements of superallowed
Fermi S decay lifetimes 1s ) s~n 8 = 0.226 0.009, and making
infra-red corrections we learn
(26)
Using Eqs. (19), this constrains nVA and nAA by
2V\ 2 -1 (27)
VA
which shrinks the allowed region to the area in a'b'c'd' in Figs. 1

and 2. The lower and upper bounds corresponding to this region
are given in Table 2. The mixing angle s becomes very tightly con-
strained, but still m2 could be as low as 3.30 times ml (at the
point c'). The relative amount of (V+A)2 given by s in Eq. (22b)
could be as large as 9.2%. Thus there is still considerable room
for a (V+A)2 interaction.
There are a number of striking consequences of the model.

Here are some of them:
1) Wrong-handed neutrinos (V R ) and antineutrinos (V L) should ex-

ist. Up to about 1% of neutrinos and antineutrinos from S decay
could be wrong-handed.
2
2) The relative amount of (V+A) grows to 1 as the momentum trans-
2
fer -q increases. The growth could be surprisingly rapid. For in-
500% near q 2 =
stance, if ml = 60 GeV and m2 = 170 GeV, s ~s
=-(150 GeV)2.
Tab Ie 2
Theory dependent bounds on weak interaction parameters
Quanti ty Lower Upper

bound bound
V-A
P(e)F/P(e)F 0.981 1.0
V-A
P(e)GT/P(e)GT input 1.0
P(e) /P(e)V-A 0.983 1.0

11 11
P(l1) /P(l1) V-A 0.985 1.0
7f 7f
P 0.750 0.750
0 0.750 0.750
t;, 0.983 1.0
n 0 0
m2/ml 3.30 00
tan 1;; -0.006 0.006

nAA 0.976 1.024
-nVA 0.823 1.012
r -0.006 0.006
s 0 0.092
3) In deep inelastic neutrino scattering, a high y anomaly can

never occur. This is because any new flavoured quarks must have
left- and right-handed couplings to any other quark with which they
couple weakly. The parton model and the effective Lagrangian in
Eq. (17) imply that the cross-section for scattering left-handed
neutrinos is
686 R. BUDNY
where q is the quark distribution and q the antiquark distribution.

For scattering on isosca1ar targets at energies above charm thresh-
old, the distributions are
and
(29)
The expression for the cross-section for right-handed antineutri-

cc - 2
nos, dcr (V R) results from moving the (l-y) factor to the first
bracket in Eq. (28). Note that there are no interference terms
proportional to r in these results since partons are massless
Also, since r2 is of order s2 and thus negligible, these results
are indistinguishable from those of the usual V-A parton model.
The cross-sections for scattering the wrong-handed v R and v L are
very small except at large q2. They are
and similarly for VL with (1_y)2 moved to the first bracket.
When the momentum transfer _q2 gets large compared with m~,
the relations for rand s in Eqs. (22a) no longer hold. Instead,
r + 0 and s + 1 giving (to lowest order) the exact V-A results for
cc cc - 2
dcr (V L ) and dcr (v R ), except for a q dependent normalization.
4) In deep inelastic muon capture of polarized muons, the reac-

tions are related to the charged current reactions discussed above
+ cc
and thus dO(~L) = do (~L) etc. This implies
dcrt~t") - dcr(tA""l!: )
=
da- l\'t') +da- (f"p. )
(3:!.)
-
and similarly for dA- with (1-y)2 moved. Dropping terms of order
~2 and using S2 = (q2 - mf)2J{q2 - m~)2 + 8(~)2 gives
(32)
These are very close to 1 when q2 is small compared to mf. At

large _q2 they vanish as parity non-conservation vanishes.
5) Although the model predicts that atomic parity violation in

bismuth could be vanishingly small, it also predicts a neutral cur-
+ - + -
rent effect in e e + ~ ~ which should be measureable at PEP and
PETRA! The effect comes from the interference of the photon with
the ZO which has jAjA coupling. It generates a cos 8 term in the
angular distribution of the ~- or ~+, whose magnitude, relative to
the usual (1 + cos 2 8) term iS 19 )
(33)
688 R,BUDNY
where m is the mass of the ZO with jAjA coupling. This could be

z
of order 10% at PEP and PETRA.
I conclude that the SU(2)L x SU(2)R x U(l) model with man~
fest left-right symmetry is viable and generates new motivations

for a number of experiments. These include (a) further measure-
ments of decays, especially precise measurements of P(e)G.T. since,
as shown in the Figures, these determine the lower bound of m2/ml;
(b) measurements of parity violation in e+e- ~ ~+~- at high ener-
gies; and (c) experiments with polarized muon beams.
REFERENCES
1) S. Weinberg, Phys. Rev. Letters 19, 1264 (1967).

A. Salam in Elementary Particle Theory, edited by N. Svartholm
(Almquist and Wiksell, Stockholm, 1968), p. 367.
An excellent review is M.A.B. Beg and A. Sirlin, Ann. Rev.
of Nuclear Sci. 24, 379 (1974).
2) R. Budny and T. Hagiwara. to be published.
3) N. Cabibbo, lecture at this School.
4) P. Baird, H. Brimicombe. E. Fortson et al., Nature 264, 528

(1976).
5) T.P. Cheng and L.F. Li, Phys. Rev. Letters 38. 381 (1977).
6) H.A.B. Beg and A. Zee, Phys. Rev. Letters 30, 675 (1973).
F.A. Wilczek, A. Zee, R.L. Kingsley and S.~ Treiman, Phys.
Rev. D12, 2768 (1975).
H. Fritz~, M. Gell-Mann and P. Minkowski, Phys. Letters 22!,
256 (1975).
A. De Rujula, H. Georgi and S.L. Glashow, Phys. Rev. ~, 3598
(1975).
7) H. Paar, lecture at this School.
8) S.L. Glashow and S. Weinberg, Phys. Rev. ~, 1958 (1977).
9) F. Gursey, lecture at this School and F. Gursey and P. Sikivie,

to be published.
10) B.W. Lee and S. Weinberg, Phys. Rev. Letters 38, 1237 (1977).
P. Langacker and G. Segre, ibid. ~, 259 (1977).
11) G. Ross and T. Weiler, preprint in preparation.
12) J.C. Pati and A. Salam, Phys. Rev. DlO, 275 (1974).
R.N. Mohapatra and J.C. Pati, Phys.~v. Dll, 566 (1975);
ibid. 2558 (1975).
G. Senjanovic and R.N. Mohapatra. Phys. Rev. D12, 1502 (1975).
R.N. Mohapatra and D.P. Sidhu, Phys. Rev. Lett;rs 38, 667 (1977).
J.C. Pati, S. Rajpoot and A. Salam, preprint ICTP/76/ll.
13) M.A.B. Beg, R.V. Budny, R. Mohapatra and A. Sirlin, Phys. Rev.
Letters 38, 1252 (1977), and to be published.
6~ R.BUDNY
14) S. Weinberg, Phys. Rev. Letters 29, 388 (1972).

A. De Rujula, H. Georgi and S.L.-Clashow, Ann. Phys.
103, 315 (1977).
15) M.A.B. Beg, R.N. Mohapatra, A. Sirlin and H.-S. Tsao,

Rockefeller preprint.
16) c.f. R. Marshak, Riazuddin and C. Ryan, Theory of Weak

Interactions in Particle Physics (Wiley, New York, 1969).
17) A.M. Sachs and A. Sirlin, in ~luon Physics, edited by V. Hughes

(Academic, New York, 1975), Vol. II, p. 49.
18) D.H. Wilkinson and D.E. Alburger, Phys. Rev. Ell, 2517 (1976).
19) R. Budny, Phys. Rev. D14, 2969 (1976).

QUARK AND LEPTON MIXING
. *)
N. Cab1bbo
Laboratoire de Physique Theorique et Hautes Energies
Universite Pierre et Harie Curie, Paris
1. INTRODUCTION AND OUTLINE
Weak interactions are known to give rise to the violation of

different flavour quantum numbers. The list of the victims include
strangeness and charm: it will probably include the quantum number
associated new quarks - the alternative being the existence of new
unconditionally stable hadrons, and it might also include quantum
numbers in the lepton sector, such as the muon number.
In gauge theories of weak interactions, the violation of

flavour quantum numbers arises from particle mixing: the particles
which belong to different multiplets of the gauge group are linear
combinations of particles which have definite mass. The first well-
established mixing effect is that associated with the violation of
strangeness. In the originalformulationl~ which predated the quark
model, this was presented as a mixing of ~S ~ 0 and ~S = 1 currents,
determined by an angle 8. In the quark language, one has a simple
descripton of this effect: the charge -1/3 quarks which are the
partners of p and c in weak isodoublets are:
*) On leave from Universita di Roma, Italy.
691
692 N. CABIBBO
n~ = n. GOs.~ .... A c:;,..,.,~

(1.1)
\. 9 =- - h S 1'"0 e -I- A Co~ e ,

leading to the doublet structure 2 )
(1. 2)
In these lectures, I will discuss possible extensions of this mix-

ing scheme to larger quark models (in particular six-quark models)
and to the lepton world. The models I will discuss are all based
on the SU(2) x U(l) gauge theory3) of weak and electromagnetic in-
teractions, under the assumption that the charged weak currents are
left-handed.
Another restriction I shall pose is that all left-handed par-

ticles should belong either to doublets or to singlets. In fact,
in any model where particles with the same mass and helicity be-
long to different representations of the gauge group4), mixing
leads to flavour violation in "neutral current" processes at order
L
G. This general rule was at work in the pre-charm days, where ne
was a doublet, A~ a singlet, and one was led to predict first-order
transitions A 7 n~+~- (or KO 7 ~ +~ - ) . In the scheme of Eq. (1.2)
these processes are naturally forbidden. The word "naturally" is
used in a technical sense:
forbidden for any value] (1. 3)

[ naturallY]
forbidden
[ of the allowable mixing
angle
If something is not "naturally" forbidden, it can be suppressed by

assuming an "unnaturally" small value for some mixing angle.
QUARK AND LEPTON MIXING 693
One would then be inclined to strengthen the above restric-

tion to: all left-handed particles are in doublets, except that I
would like to discuss an "unnatural" model of the lepton sector,
the "economy model", to show that it can be excluded on the basis
of a combination of theoretical and experimental arguments 5)
Although in these lectures I shall discuss the lepton sector

and the hadron sector separately, the two are linked - even in the
simple SU(2) x U(l) theory - by the requirement of the vanishing of
Adler anomalies. With the restrictions to singlets and doublets,
6
this leads to ):
Number of lepton doublets Number of quark doublets (1.4)
where a lepton doublet has charges (0, -1) [example: (v , e-)] and
e
a quark doublet charges (2/3, -1/3) (example: (p, n)), in three
colours.
With the discovery of a new lepton T and a probability of a

new quark (in the Y resonance), the standard scheme [Eq. (1.2) ] can
be "naturally" extended to a three-lepton doublets/three-quark
doublet scheme. This requires the existence of still another quark -
and the corresponding family of q-q states - and another neutral lep-
ton, v T This will be called the "standard six-quark/six-lepton
model".
An alternative could seem to be the "economy model", where the

Y-associated quark and the T are ~n singlets. We will, however,
show that this model can be excluded. As a consequence, the stand-
ard 6q-6 model is the most economical model compatible with the
presently known experimental facts.
In the next section I shall discuss mixing ~n the quark sector.

The main result in this section is the fact that mixing in a s~x
quark model can lead to CP violation 7 ) . The theory of CP violation

thus obtained is natural, in the sense that we do need a small para-
meter to explain the small breaking observed in the K{-K~ system,
694 N. CABIBBO
and it agrees at the present level of accuracy with the superweak

theory7,s) in that ~ + 2n arises mainly from ~-KS mixing, while
the neutron e.d.m. is expected to be exceedingly small.
I shall then discuss mixing effects in the lepton sector.

These effects fall into two general categories:
1) Muon number violation: ~ + ey and related processes.

2) Neutrino oscillations.
In the frame of a six-lepton scheme, effects of the first kind are

possible (at detectable rates) only if the v, is heavy (I will not
discuss the possibility, proposed by Bjorken and Weinberg, of ob-
taining muon number violation through the Higgs couplings 9 )). The
main result here is that, within the six-lepton model, the anoma-
lous capture ~ + N + e + N is found to give a better test of muon
number conservation than ~ + ey. A separate section will be de-
voted to the killing of the economy model.
Neutrino oscillations are possible if the neutrinos have non-

degenerate masses. The main new result I will present is the pos-
sibility that mixing between three light (but non-degenerate) neu-
trinos leads to CP violating neutrino oscillations 10 ).
The world of lepton mixing can be full of surprises and of new

phenomena. Unfortunately, it can be totally empty: it suffices
that the three neutrinos be massless, and processes 1) and 2) above
would be excluded. Their possibility poses, however, exciting
challenges to the experimental community.
2. MIXING IN THE QUARK SECTOR
2.1 General discussion - possible CP violation
In a model which contains an equal number n of charge 2/3 (p-

type) and charge -1/3 (N-type) quarks arranged in n left-handed
doublets, the general doublet structure is
d\ - ( 4= 1, ... )", (2.1)
Here P., N. are the fields with definite mass (p, c, t ... , n,
~ ~
A, b, ) and Aik is an n x n unitary matrix
(2.2)
The charged weak current is
~ A
J+ = ~ 1>....). ~~')I'l
'V
). ,
L\C.
\.K
2. "- (2.3)
The neutral weak current is
(2.4)
where 8w is the Weinberg angle, Jem is the electromagnetic current,
(2.5)
By using Eq. (2.2) we rewrite this as
(2.6)
(2.4) and (2.6) show that the weak neutral current is diagonal for
any allowed (i.e., unitary) form of the mixing matrix A - the
theory is "natural".
696 N. CABIBBO
It is worth while recalling that (2.2) follows from the re-

+
quirement that 1-, 1 3 generates the algebra of weak isospin, and is,
therefore, essential in a gauge theory.
We note a certain degree of arbitrariness 1n the definition of

Aik : the transformation
(2.7)
1S without physical consequence, S1nce I can reabsorb it by rede-

fining the fields as
(2.8)
Such a redefinition has no effect on quark-gluon and quark-photon

interactions which are diagonal in flavour. It follows that A and
A' are physically identical if related as in Eq. (2.7). In other
terms, the matrix A is determined up to (2n-l) arbitrary phases.
It is known that a necessary condition for the violation of time re-
versal is the presence of complex couplings, in our case of complex
elements of A; given the arbitrariety just discussed, we conclude
that time reversal is conserved if A is equivalent [through (2.7)J
to a real unitary (i.e., an orthogonal) matrix A'. This is always
true for unitary 2 x 2 matrices, but not for unitary matrices of
higher rank.
2
For n 3, A depends on n =9 parameters. Subtracting the
(2n-l) = 5 irrelevant phases, we remain with four essential para-
meters, one more than those which determine an orthogonal 3 x 3
matrix. The extra parameter has the nature of an uneliminable phase.
The general mixing scheme In the six-quark model will give rise to
T (and CP) violation 7 ) .
Following Kobayashi and Maksava, we can parametrize the gene-

ral form of A by three angles: 8), 8 2 , 8 3 , and a phase parameter, 6
n A b
P C) -SlC 3 -S)S3
i6 i6
c SlC 2 C)C2 C3 - S2 S 3e C)C 2 S 3 + S2 C3e
A (2.9)
i6 i6
t SlC2 C)S2 C3 + C2 S 3 e C1 S 2 S 3 - C2 C3e
where
(2.10)
and 6 is a phase leading to CP violation.
We note that A can be made real if:
a) Any of the S. or the C. are equal to zero.

l l
b) There is a degeneration between charge 2/3 quarks or between
charge -1/3 quarks.
The first statement is easily checked by inspection: if one of the

C. or S. vanishes, A can be made real by mUltiplying rows and co-
l l +' 6
lumns by appropriate e- l factors. It follows that any CP violating
amplitude will be proportional to sin 6 n.l S.C
l l
..
The second statement is true because, if two charge 2/3 (or

charge -1/3) quarks are degenerate, we can enlarge the group of
transformations in Eq. (2.7) to include SU(2) transformations be-
tween degenerate quarks.
698 N. CABIBBO
Among the angles e., el corresponds to the original e (Eq.

~
1.1. e 3 is already known to be small, since it will lead to a

violation of the weak interaction universalityl):
(2.11)
On this basis, Ellis et al. eive the limit
s,= <. O.OE> (2.12)
A limit on S2 can be obtained by noting that production of charm by

neutrinos on valence quarks (large x) is given by
(2.13)
Since this is known to be in rough agreement with the predictions l1 )

obtained for e2 = 0, we already know that C2 is not too different
from 1, i.e., S2 is not very large. It will be interesting to re-
turn to this question in a more quantitative way.
2.2 Production and decay of B particles l2 )
If we assume the Y to be a bound state of bb, B particles

(bP, b~, bS) can have a mass of ~ 5 GeV. Their preferred decay
mechanism should be given by
b --'" C ... ,t.-+
b--"'c. ... q+q

'?t .t= p,e
, (2.14)
where ql, q are non-charmed quarks.

Channels which do not contain
a c quark would be suppressed by a sin 2 e factor. The semileptonic
branching ratio is expected to be similar to that for charm: lO~13%
for each (~v

~
or ev e ) channel.
Similarly, the preferred mechanism for neutrino production

would be off a c quark:
(2.15)
A complete event of b production would eive rise to a complex cas-

cade:
v~ -.- p ~ C
VV+!++Y.~
"e+e +X (2.16)
'X .
(-
v.,.~~- . . X,
\5e~Q"" ... 'X.
X
where ~ indicates normal hadrons, C a charmed particle. Such an
event could have up to four muons or three muons with signatures
-++ or -+-. Assuming equal semileptonic branching for Band C,
B ~ 1/10 we would expect:
~\iX
~1:2.:~ (2.17)
10
One would thus have roughly one 4~ event for each 30 events, or less
if alternative mechanisms for 3~ events exist (e.g., associated
charm production).
2.3 CP breaking
Because of the arguments given above, any CP breaking ampli-

tude will be proportional to a product of many mixing angles. One
can thus obtain a violation of the observed order, 10- 3 , without in-
troducing explicitly a very small parameter.
700 N. CABIBBO
The situation has been exhaustively discussed in the papers
of Ref. 7. The general conclusion is that the pattern of breaking

of CP predicted here is practically identical to that given by the
superweak theory, and experimentally indistinguishable from it:
i) CP breaking in the Ko-Ko system arises mainly from Ko-Ko mix-

ing: one thus finds up to very small corrections the same
predictions of the superweak theory which are supported by
the experimental data.
One obtains in particular:
(2.18)
so that a 10- 3 violation can be obtained if S2 ~ S3 ~ 6 ~ 1/10.
The same combination appears in the CP violation ~n 6S = 1 de-

cay, so that
ii) CP violation in 6S = 1 decay, apart from Ko-Ko mixing effects,

will not be observable.
Finally, this theory agrees with the superweak one ~n predict-

ing that
iii) The e.d.m. of the neutron will be very small.
The reason for this result is that the first order (one W ex-
change) can easily be shown to be CP and T conserving. This is true
for any amplitude (including the neutron-photon vertex) which con-
serves all flavour quantum numbers. In a diagram for such an ampli-
+
tude, where the W ~s emitted by a (qlq2) term in J , it must be re-
absorbed by the corresponding (q2q 1) term in J
- The diagram will
then be proportional to A A* ,which is real.
qlq2 qlq2
Maiani and Ellis et al. (Refs. 7 and 8) have given an evalua-
tion of the second-order weak contribution, which leads to
The uncertainty in the result being linked to the ignorance of the

mixing parameters and the masses of band t quarks.
3. LEPTON NIXING
3.1 Outline of the section
In this section we review possible effects of lepton mixing

~nthe left-hand models. These models can be ranged in possible
classes according to the mass-spectrum of the neutrinos, which de-
termines the possible range of new physical phenomena, such as w-e
transitions (Section 3.2) and neutrino oscillations (Section 3.5).
Huon-electron transitions are possible if the new neutrino is heavy
(Section 3.4). Neutrino oscillation can arise between light, but
non-degenerate, neutrinos. If the three neutrinos are light and
non-degenerate, neutrino oscillations can also exhibit CP and T
violation phenomena (Section 3.5). The possibility of having only
two left-handed neutrinos is excluded (Section 3.3). Finally, and
unfortunately, the possibility subsists of having three massless
neutrinos which would not give rise to either type of new physical
phenomena.
The range of experimental possibilities ~s very large: the

discovery of new phenomena or the setting of new limits on their
existence.
3.2 Review of w-e transitions
Huon number violation effects can appear ~n gauge theories, as

a consequence of lepton mixing.
The three processes which have been extensively studied are:

702 N. CABIBBO
p~e-+"'{
~ -. 3e.
(3.1)
\l-... N -+ e"," N
1 3)
The absence of ~ + ey at the value predicted by Feinberg on the
basis of the one-neutrino model prompted a set of theoretical in-
vestigations by Feinberg and Weinberg 14 ) , and by Gatto and myself 1S )
in which the merits of the three processes were compared, and dif-
ferent theoretical models were considered. We were led to conclude
that the only natural way to forbid the three processes was the
existence of a muon-electron symmetry which could only be realized
16)
through the existence of two separate neutrinos ,one coupled to
the muon, the other to the electron. This view was, as we now know,
correct. The advent and general acceptance of gauge theories has
allowed a sharpening of the theoretical predictions. The process
~ + ey can arise in some models at second order in perturbation
theory. The resulting amplitude is always computable in finite
terms, since in a renormalizable theory the only divergences are
those which correspond to terms in the Lagrangian, i.e., to possible
counter-terms, and gauge invariance forbids a direct ~ + ey coupling.
It has, in fact, been proved that any ~ + ey or ~-e (mass mixing)
term can be eliminated from a Lagrangian with minimal (renormali-
1 7)
zable) interactions by a redefinition of ~ and e fields .
For the other two processes, the situation is more complex and
we have to distinguish between two classes of models. In the first
class are models where lepton mixing can give rise to a ~-e-z coupl-
ing term in the Lagrangian. In these models ~ + 3e and the anoma-
lous capture process ~ + N + e + N can appear through a first-order
weak interaction:
j.l e I e
I
I
I
z Iz
I
(3.2)
I
I
I
I
e --------~-------- e q q
An example of class I will be discussed in the next subsection.

This model can be excluded by a combination of theoretical and ex-
perimental arguments in a conclusive way. This is not true of other
models of class I, but models of this kind are theoretically un-
favoured because the small experimental limit on these processes
would require an unnaturally small value for some combination of
mixing angles.
In class II are models where a j.l-e-Z coupling is naturally for-

bidden. In some of these models j.l + 3e and j.l + N + e + N can still
appear as a second-order weak effect. To this class belongs the
standard six-lepton model, to be discussed in Section 3.4. Other
models of this class have been discussed in recent literature.
The experimental limits on processes (3.1) are at present as

follows (at the time of the School)
B", _ < (Ref. 18) (3.3)
l.1 3e. 1.9 10- 9

'Re :-
~
< 'lC. (Ref. 19) (3.4)
p .. all
\,+N _ e.. N (on Cu,

10- 8 (3.5)
RN = < 1.b 'I(.
Ref. 20)
f-i+N ... ~. N'
704 N. CABIBBO
The experimental situation ~s in rapid development due to a new

series of experiments now under way at the new meson factories.
Two new results have recently appeared:
(Ref. 21) (3.6)
(on S,
(3.7)
Ref. 22)
and more are expected.
The results (3.4), (3.5) and (3.7) can be used to give limits
on a possible direct W-e-Z coupling. If we assume
(3.8)
we find, for the anomalous captureS)
-= (3.9)
The factor on the right-hand side of this expression arises as fol-

lows: a numerator Z2F~, where FN ~s the nuclear form factor at
q2 = m~, s~nce the anomalous capture can occur coherently on the
nucleus, and a denominator ZC(l + 3gi) , where C is a factor 23 ) which
takes into account the Pauli suppression of the normal capture
W + P + V + N due to the fact that part of the phase space for the
W
final neutron is already occupied. The over-all effect is an en-
hancement of the anomalous capture by a factor of ten:
IV 10
(3.10)
We thus obtain from (3.5) for sin 8W 0.3 a bound
(3.11)
and from (3.7)
(3.12)
The experimental limit on ~ + 3e gives a poorer bound, due to the

lack of an enhancement factor (3.10)
(3.13)
3.3 The economy model
In the standard left-handed model, the existence of a new lep-

ton, T-, implies the existence of an accompanying neutrino, v T '
This model will be discussed in the following sections. In this
section we show that one extra neutrino is necessary, by excluding
the possibility of a model - the economy model - with only two left-
handed neutrinos. The new neutrino could be left-handed - as in the
standard model - or right-handed, a possibility that we will not dis-
cuss here.
In the economy model, the new lepton is a singlet, up to mixing

of its left-handed components with those of e , ~. This mixing is,
~n fact, necessary, for T to be unstable. The most general mixing
scheme involves two mixing parameters: a and S. We have two
doublets:
706 N. CABIBBO
Ve.] 1.
2.
[('\-~')e ... ~"t-.)
( (1. ~'Z.) P- -t "("t- - \1'( e-.) vp. ]

(3.14)
and one left-handed singlet
(3.15)
while all R fields are singlets. In the above expressions we have

neglected terms higher than the second in a and S. The presence
of an e- component in the muon multiplet implies a direct ~-e-Z
coupling:
, (3.16)
In the standard model this term would be cancelled by a similar

term arising from the l doublet. In the present model this term
survives, leading to a limit on Sy [compare (3.16) with (3.8) and
(3.11)/(3.l3J
On the other hand, one easily computes the total rate for l decay,
which is proportional to the sum of the two mixings
(3.18)
-1 -9
We may safely assume rl < 0.3 10 s, which would give a mean
free path of ~ 10 cm in the SPEAR or DORIS experiments. If we
take sin 2 8W = 0.38, which maximizes the right-hand side in

Eq. (3.18), we have then:
-4
'3 )\. 10 . (3.19)
The two bounds (3.17) and (3.19) together imply that either
of"
)
(3.20)
In either case we obtain detailed predictions for the branching

ratios in T decays, which disagree with the known results. In
particular, we obtain
-r
q.
(3.21)
while experiment indicates equal branching ratios. Another area

of violent disagreement is 1n the prediction of sizeable decays in-
to three charged leptons with branching ratios given in the follow-
ing table:
S2 y2 y2 S2
- + -
~ ~ ~ 'V 0 J%
-
e e e
+ -
3% 'V 0
- + -
~ e e 'V 0 2%
e
- ~ +~ - 2% 'V 0
708 N. CABIBBQ
The existence of these decay modes at the predicted branching ra-

tios seems to be excluded: the economy model is rejected; one new
neutrino, possibly left-handed, is necessary.
3.3 Muon electron transitions in the

six-lepton left-handed model
If the neutrinos are massless, we have the multiplet scheme
CJ l..
(3.22)
In this case, neutrino (or charged lepton) mixing would not have
any physical consequence. Huon electron transitions are possible
if neutrinos are non-degenerate and mixed, but the corresponding
amplitudes will be proportional to mass differences, so that ap-
preciable rates are possible only if V is much heavier than V ,V
T e ]J
In this situation the result will essentially be independent from
the mass of the light neutrinos, so that we can assume them to be
massless.
The mixing scheme involves two parameters, a and b, and is

specified by the following doublets, which we write neglecting
third-order terms in a and b:
)
e-) L
[c 1- ~) -vp. + DV~ - ab ve.> p-] L (3.23)
rl l1-S~- ~)V1i -ave. -b"tI-.> -c- J

2. 2. L
We note that the weak neutral current is now diagonal, i.e.,

we have no direct ~-e-Z coupling.
If we assume MVT ~ M (as will be required in order to obtain

T
appreciable rates for ~-e transitions) the T decays through V and
e
V emission, and we are led to:
~
(3.24)
1 -9
Assuming, as in Section 3.3, a lifetime r~ < 0.3 10 we have a
(3.25)
This limit can probably be improved by a factor of ten through an

analysis of the SPEAR and DORIS results.
We note that V (if heavy) is not emitted ~n normal weak in-

T
teractions, which would correspond to the effective transitions:
(3.26)
'Z.
r'\-E
\ 2
) -VP. - abv~
The two neutrino states are not orthogonal, and not equally nor-
malized, so that:
i) The ratio TI ~ eV/TI ~ ~V is different from the standard pre-

diction:
1t,~ev _ (1+b2._o..2.)
1t. .... \A',) (3 27)
ii) Neutrinos from TI ~ ~v, K ~ ~v can appear as electron neutri-
nos with the probability a2b2

710 N. CABIBBO
-\)+w-.e--+
-.:>.. N~ p-.,. ... (3.28)
The experimental value for n + ev/n + ~v IS in slight disagreement

24)
WIth the standard result , leading to:
(3.29)
This result is compatible with b 2 - a2 0, and can be interpreted

as a bound on Ib 2 - a2 1.
25)
Experimental results from Gargamelle on the ratio in
Eq. (3.28) lead to:
(3.30)
The transition ~ + ey arises from the diagram
ve , VT
~
0 I
I
e
(3.31)
W I
I
I
I
I
Y
26)
One fInds
1<1=
Inserting the limit (3.30), and neglecting corrections of order

(M2 nC) 2, we have
vT -W
(3.32)
27) . .
We h ave made a study of the relatIve merIts of ~ ~ ey and the
other two processes as a test of muon number violation. Within the
model under discussion, Fig. 1 shows the values of the three ratios
as a function of Mv = m, assuming sin 2 8W = 0.33 and (a b)2
-3 T
= 2 x 10 (equal to the limit in Eq. (3.30. In this computation
we have treated exactly the dependence on Mv
T
10- 6= - - - - - - - - - - - - - , Our results show that in left-
handed models, the anomalous cap-
ture is (for moderate values of
MVT ) 20 times more sensitive than
~ ~ ey. The situation is differ-
27
ent ) in other models including
right-handed currents, where ~
and R are of comparable magnitude.

y
This means that a possible detec-
tion of the two processes would
give precious information on the
structure of the theory. The rate
predicted for ~ + 3e is, In any
case, lower than that for ~ + ey,
so that this experiment looks pro-
mising.
Fig. 1
712 N. CABIBBO
In conclusion, the present theoretical situation suggests the

necessity of further searches for both ~ ~ ey and ~ + N ~ e + N,
and of an improvement of the experimental situation on the TI ~ eV
branching ratio as well as on the ratio V ~ e/v ~ ~.
3.5 Neutrino oscillations: CP and T violation
It has been observed by Pontecorv0 2S ) that if v and V~ are

not both massless and non-degenerate, they can be mixed - in ana-
logy with the n-A mixing responsible for the decay of strange par-
ticles. If Nl and N2 are the mass eigenstates, we can write:
v\'- = Nt c.os-e~ -t- N2. sWIG"

"e, =- N\ SWIG" -t- N2 c.o~ e~
(3.33)
A neutrino of momentum p, initially produced as V (example: TI

+ ~
~
+
~ ~ V ) can then, at a distance from the source, appear as aVe'
~
with a probability which is easily found to be
(3.34)
This phenomenon can be detected through the appearance of V events

in a V beam, or through an apparent lack of v. We have, in fact,
~ ~
(3.35)
As was the case in the four-quark model (Section 2.1), one easily
sees that (3.33) gives the most general mixing among two neutrino
species, i.e., that any mixing involving phase factors can be re-
duced to the real form (3.33). If we have three neutrino species,
non-degenerate in mass, the mixing scheme may include phase factors,

just as in the six-quark model discussed in Section 2.1, and give
rise to violations of time reversal, T and of CplO).
To see how these violations may be manifested, it is conveni-

ent to treat the oscillation phenomena in a general way, valid for
any number n of neutrino species. We will assume the species to
be of light mass, so that different species can be produced co-
herently in decay processes as TI + ~v. We will call N. the mass
l.
eigenstates (i 1, n), mi the mass values, and v the field
=
which diagonalize weak interactions: {v} =

{v' v~, ... }. The
two sets will be related by a unitary transformation:
(3.36)
(3.37)
We recall from the discussion in Section 2.1 that if n =3 and

there is no mass degeneracy, Aik can contain essential phase fac-
tors. It is easily seen that for n > 3 the condition for the pos-
sibility of non-eliminable phases is the existence of three or
more non-degenerate mass values.
Assume we produce a v (for instance V~) at time t = 0 with

momentum p. The time development of this state is computed by
using Eq. (3.36):
(3.38)
where
'Z..
-
m",
(3.39)
714 N. CABIBBO
We can now use Eq. (3.37) to obtain
(3.40)
It is interesting to define transition amplitudes
(3.41)
whose norm gives the transition probability:
(3.42)
We note that (from Eq. (3.41)):

~
(3.43)
time reversal ~s equivalent to exchanging the initial and final

states. Time reversal invariance would then require
(T conserved) (3.44)
This is certainly true in "diagonal oscillations", Le., when 9- = i.

We have the (trivial) statement: "diagonal oscillations conserve T".
For i f 9-, however, (3.44) may be invalid if A contains essential
phase factors. Let me give an extreme example: assume
1 1 1
(3.45)
A= 1- 1 Q 0.. t:l -
-
21r .
e.-,cp -3 '\.
~ tt
1 0.. 0..
This matrix has the interesting property that the three neutrinos
are maximally mixed (this isnot possible for three neutrinos with
a real matrix 29 )).
Let me also assume that the three masses are equidistant:

m~ - m~ = m~ - mi = m2 From (3.40) one easily obtains:
1
=-
(3.46)
The presence of a term which is not even in t ensures, through

Eq. (3.43), that T (Eq. (3.44)) will be badly violated in this ex-
ample. At t = 4np/3m 2 we would have P(V~ + v) = 0,
P(V + V~) 1. The CPT theorem implies that a T violation goes
together with a CP violation. In the case of interest, CPT implies
that the oscillations of right-handed antineutrinos will be related
to those of left-handed neutrinos by:
(CPT) (3.47)
A violation of T (Eq. (3.43)) entails a violation of CP, which would

require:
(CP) (3.48)
In conclusion, the existence of more than two neutrinos enriches

the scope of neutrino oscillation experiments. If oscillations are
detected, one might wish to check whether they are T and CP conser-
ving. A test of CP requires the comparison of non-diagonal neutrino
and antineutrino oscillations, e.g.:
716 N. CABIBBO
-VtJ- ~ V-e VS. VIJ-~Ve )

0'-
(3.49)
\J~_ v-r; VS, 'V.... ~ ~"t
The most convenient test for T violation seems to be the search for
odd (sin m2 /p t) terms ~n the oscillation probability.
The phenomenology of neutrino anomalies is possibly very rich.

Apart from oscillations, it can contain neutrino instability, pro-
30)
posed some time ago in the frame of the solar neutr~no problem ,
and possible non-orthogonality effects, discussed in the previous
section. It is left to the ingenuity of the experimenters to find
if some of these potential riches are used by nature.
REFERENCES
1) N. Cabibbo, Phys. Rev. Letters 10 (1963) 531.
2) Y. Hara, Phys. Rev. 134 (1964) B701. The necessity of this

structure was established by S. G1ashow, J. I1iopou1os
and L. Maiani, Phys. Rev. D2 (1970) 1285.
3) A. Salam, in "Elementary Particle Physics", ed. N. Svartho1m

(Almquist and Wiske11s, Stockholm, 1968).
S. Weinberg, Phys. Rev. Letters 19 (1967) 1264.
4) S. G1ashow and S. Weinberg, Harvard preprint HUTP - 76/A158

(1976).
5) G. Altarelli, N. Cabibbo, L. Maiani and R. Petronizio, Phys.

Letters 67B (1977) 463.
D. Horn and G. Ross, Phys. Letters 67B (1977) 460.
6) C. Bouchiat, J. Iliopou1os and P. Meyer, Phys. Letters 38B

(1972) 519.
7) M. Kobayashi and K. Maskava, Progr. Theor. Phys. 49 (1973) 652.

L. Maiani, Phys. Letters 68B (1976) 183.
S. Pakvasa and H. Sugawara, Phys. Rev. D14 (1976) 305.
8) J. Ellis, M.K. Gaillard and D.V. Nanopoulos, Nuclear Phys. B109

(1976) 213, and papers in Ref. 7).
9) T.D. Lee, Phys. Rev. D8 (1973) 1226.

S. Weinberg, Phys. Rev. Letters 37 (1976) 657.
10) N. Cabibbo, Phys. Letters 72B (1978) 33.
11) See, for example, G. A1tare11i, N. Cabibbo, L. Maiani and

R. Petronzio, Phys. Letters 48B (1974) 435.
12) An extensive treatment has recently been given by J. Ellis,

M.K. Gaillard, D.V. Nanopou1os and S. Rudaz, Nuclear Phys.
B131 (1978) 285.
13) G. Feinberg, Phys. Rev. 110 (1958) 1482.
14) G. Feinberg and S. Weinberg, Phys. Rev. Letters 3 (1959) 111;

ibid. (E) 3 (1959) 244.
15) N. Cabibbo and R. Gatto, Phys. Rev. 116 (1959) 1134.

718 N.CABIBBO
16) N. Cabibbo and R. Gatto, Phys. Rev. Letters 5 (1960) 116.

Two neutrino theories had been proposed before, e.g.,
B. Pontecorvo, J. Expt1. Theor. Phys. (USSR) 37 (1951) 1751.
J. Schwinger, Ann. Phys. 2 (1957) 407.
17) N. Cabibbo, R. Gatto and C. Zemach, Nuovo Cimento XVI (1960)

168. Partial proofs were given in Ref. 16) by G. Feinberg,
P. Kabir and S. Weinberg, Phys. Rev. Letters 3 (1959) 527.
18) S. Parker, H. Anderson and C. Rey, Phys. Rev. 133 (1964) B768.
19) S.M. Korenchenko et a1., Sov. Phys. JETP 43 (1976) 1.
20) D.A. Bryman, M. Blecher, K. Gotow and R.J. Powers, Phys. Rev.
Letters 28 (1972) 1469.
21) P. Depommier et a1., Phys. Rev. Letters 39 (1977) 1113.
22) A. Badertscher et a1., Phys. Rev. Letters 39 (1977) 1385.
23) A. Primakoff, Revs. Mod. Phys. 31 (1959) 802.

J.C. Sens, R.A. Swanson, V.L. Te1egdi and D.D. Yovanovitch,
Phys. Rev. 107 (1957) 1464.
24) D.A. Bryman and C. Picciotto, Phys. Rev. D11 (1975) 1337.
25) E. Bellotti, D. Cavalli, E. Fiorini and M. Ro11ier, Nuovo

Cimento Letters 17 (1976) 553.
26) B.W. Lee, S. Pakvasa, R.E. Schrock, H. Sugawara, Phys. Rev.

Letters 38 (1977) 937.
27) G. A1tare11i, L. Bau1ieu, N. Cabibbo, L. Maiani and R. Petronzio,

Nuclear Phys. B125 (1977) 285. W.J. Marciano and A.I. Sanda,
Phys. Rev. Letters 38 (1977) 1512.
1512.
28) B. Pontecorvo, JETP 53 (1967) 1717.
29) S. Nussinov, Phys. Letters 63B (1976) 201.
30) J.N. Bahca11, N. Cabibbo and A. Yahi11, Phys. Rev. Letters

28 (1972) 316.
DIS C U S S ION
CHAIIDIAN: Prof. N. Cabibbo
Scientific Secretaries: L. Baulieu, M. Dine, G. Valent
SANDA:
I would like to ask a philosophical question concerning which

I would appreciate your wisdom. In this school there has been
much discussion about the relation between MZ and MW'
In the most
naive version of the Weinberg-Salam model with the simplest Higgs
structure, me 1'W- .5 MeV is related to MW ~ 65 GeV. Should this be
a source of concern in a natural theory?
CABIBBO:
This problem was studied a few years ago be Georgi, Weinberg,

and others. They proposed that the ratios of small to large masses
could be computed if one had a model such that:
1) in lowest order in 0( "small" masses were zero,
2) in higher orders one had computable values for small masses.
In such a model one could obtain an expression for the electron mass
of the form, m =c(~~f where f is a function of other parameters
e
of the theory. Unfortunately, the explicit examples of this mech-
anism which have been produced up to now involve many arbitrary
parameters (for example the masses of many W bosons). However, the
whole scheme is very attractive, and its study should be pursued.
719
720 DISCUSSION
FISHER:
In the model you discussed this morning, do you expect larger

1
or sma11 er CP v~o
at~ng e ff ects ~n
. teD
h an ~n t h e KO-Ko
D system t h
0-0
system? Could such effects be detected in e+e- annihilation?
CABIBBO:
0-0 0-0
As in the K K system, CP violation can appear in D D mixing,
possibly with a strength comparable to that in KOK o mixing. However,
in contrast to the KOK o system, mixing effects are expected to be
0-0
very swall in the D D system since the leading decay channels
o -0
(having As =Ac = +1) will be separate for D and D. It is only
rare (proportional to sin 26 ) channels which are common to both
o -0 0-0
D and D decays and can lead to D D mixing.
One still expects the two states Dl and D2 to have definite
1.="'1 and (~I-"""')-, to

lifetimes and masses. However, in view of the above arguments, one
expects both be small, compared to the average
lifetime (the ratio being of order sin 2 B). This, together with the
expected smallness of the DO and nO lifetimes will make it very
difficult to detect Dl -D 2 oscillation and CP violation effects.
LIPKIN:
One way to observe CP violation would be to observe events

where e +e - .. D1 -D2 and both decay into the same channel (for example
W+1r-) thus violating CPo
CABIBBO:
Yes, this would give a test but the expected combined branching
ratio would be sin~~ x (CP violating factor). This would be very
+ -
difficult to identify against e e ~4-r background!
DISCUSSION 721
DE LA TORRE:
You have shown us one way to introduce CP violation in the

standard model. This involves the introduction of 3 Cabibbo-type
angles and one phase. My question is: are there further constraints
which reduce this rather large number of free parameters?
CABIBBO:
L 1.. 2-
Not many. In the 3x3 matrix we introduced, the sum ~+S, Co 3 ::.
'1. '1. 1. 'L
'-I .... $, - $, S~ must be equal to 1.00 .05 in order to be
compatible with the data which justifies the old Cabibbo theory.
This doesn't put very strong constraints on S) however. It will
also be possible to measure the matrix elements which give the ci\
and cn couplings, but it will be a long time before this is done.
Of course we also have the relation E~SISl S~) where ~ ~s the
CP violating parameter of the K K system. If we believe the model
we expect the new angles to be comparable to the old one, perhaps
smaller.
PAAR:
You can see from the 3x3 matrix that it will be very hard to
make band t quarks because the production of valence quarks is
proportional to two sin~ type factors. So it will be very hard
to see band t quark production by neutrinos if this model is
correct.
CABIBBO:
This is essentially correct. The only way to produce b or t

quarks without Cabibbo suppression is from the sea. For example,
these could be produced through the reaction ~ t'f.."i-hI:. Of
course, the value of the threshold for this reaction will be rather
high for present neutrino experiments. Let me recall that the b
722 DISCUSSION
would be a perfect source of rare trimuon events and same sign

dimuons which would have very spectacular signatures even if their
rates are substantially smaller than that for charm production.
CELHASTER:
Could you describe the process being measured in the atomic

physics experiments and explain the predictions of the standard
model?
CAEIBBO:
Consider a schematic of the level structure of the atom. It

has a level structure of the form:
The transition between the two S states would be M~ if parity is

conserved. Parity violation leads to a mixing of the Sand P states
so that the transition between the two S states will be MI + 1 EI.
By sending in a linearly polarised photon and measuring the rotation
of the photon polarisation we can determine the factor ~.
In the standard model the electron neutral current is not pure
vector, and in the non-relativistic limit the effective Hamiltonian
contains a piece proportional to (ei(? ;)e) .,....p which will give
rise to parity violating effects (S - P mixing).
WEILER:
I would like to make a comment on parity violation in Bi.

Experimentally one excites a metastable state in atomic Bi with
linearly polarised light and measures the rotation of the polar-
isation of the emitted photon. Because of the very short range of
the weak interaction, the effect is proportional to Itf'lo)1 ~ This
quantity has been computed by several theorists. Extreme assumptions
DISCUSSION 723
lead to variations in the predicted value by no more than a factor

of two. Two independent experiments have obtained results which
are consistent with parity conservation and sit 4-8 standard devia-
tions from the predictions of the standard model.
REITHLER:
The negative results of searches for parity violation in atoms

exclude the standard Weinberg-Salam model. This model must be ex-
tended. Could you say something about the relations among neutral
current cross sections predicted by these models?
CABIBBO:
If you believe the atomic physics data, there are many ways to
construct theories without parity violation in atoms. But you have
to keep in mind that the neutral currents observed in neutrino
reactions (either Jj.N or ~e ) agree quite well with the predict-
ions of the Weinberg-Salam model (in particular, the leptonic
neutral current seems not to be pure vector). If you accept the
atomic physics result and you wish to remain in the SU(2) X U(l)
framework, you can add a right handed doublet, (~)~ where N is
a neutral heavy lepton. This would give a purely vector neutral
current for the electron,and no parity violation in atoms (or to
a small violation, compatible with present limits). An alternative
to this approach is to enlarge the gauge group. Salam has recently
proposed an SU(2) X SU(2) X U(l) model which predicts no parity
violation in atoms but does predict parity violation in neutrino
induced neutral currents.
I wish to say that the point of my lectures was to illustrate
a possible model, not to show the absolute truth. The atomic
physics experiment will tell us something about physics, but what-
ever the results are it won't be a crisis for gauge theories if the
simplest model has to be modified.
724 DISCUSSION
BUDNY:
The predictions of the Weinberg-Salam model for atomic parity

violation depend on the Weinberg angle. The effect decreases as
sin 29'w decreases. Host of the quoted results assume sin2~ =
0.35, but the CERN-SPS neutrino data give sin2~-,= 0.25. Does this
reconcile the predictions with the measurement?
REITHLER:
A change in the value of sin1rMf affects only the vector part

of the nuclear weak current which is proportional to Q = (1-4
sin ~~ ) Z-N (Z and N are the charge and the atomic number of the
nucleus) and doesn't modify the axial part of the electronic neutral
current. Taking sin2sr~ .25 instead of 0.35 only modifies the
theoretical prediction of parity violation by 25%, which is not
enough to resolve the discrepancy.
REITllLER:
From the model discussed, an upper bound for the decay jt;er
has been predicted. How large is the predicted bound for the
analogous decay of the muon neutrino ?
Answer by BAULIEU and SANDA.
This question has been asked many times, and it is worthwhile

to give a precise answer. ~-+.,:< where Y
Let us take a decay
~s some massless, spin 1/2 object (not necessarily Ye ).
The matrix element for the decay is
mt r,r ~). 'Y(P:a.J u).r [A + B ~5 J~L~l)

where
'=rl-~~ I ~ - {[~/~J
DISCUSSION 725
this leads to
i;
where m is he mass of ~. So any experimental number for A and B
is interesting. How is A and B related to more fundamental para-
meters of the weak interaction theory? A and B can be computed in
a rather general way within the context of gauge theories, Consider
a Lagrangian
~ =- 4/
t!'
.2. ~,\ ~ ~ (A:j y.. . . .!. ij ... ) Lj
l j
C" = '!!)
+ Higgs contributions &
where >'i' stands for Y and ~ J L.is a set of heavy leptons with
mass ".... and charge Q which co~ple to both Y and YJot. We assume
')
that '\Nl<.<~ .
~J
M.~.#
~
where M~ is
~
the mass of the \Iboson.
A and Bare:
r
if the second term is non-vanishing
- 3D( (Mil"r)2.(~)1
r~~V'( -;r ~ Q r~
'2.
e>'Y lz.J
'I( L, ('t.1J bIJ'J ~)2-+(L

J
* . IA,.~ W J
b.lJ 5)7
~
~
2
Note that this requires a left-right mixing. Although it is

possible to construct a gauge model for which (2) is applicable it
will not be very attractive. FrODI the experimentalist I s point of
view it should not be ignored.
In some theories,
J
or .q'=Oor
J J J J
b, "tLl':'O. L "I' b, In this
case the first term in (1) may contribute and thus
726 DISCUSSION
This formula is applicable for theories in which there is no leptonic

G.I.M. cancellation.
Now, in realistic models one must also incorporate )t-; "if
-
~
branching ratio ~ 10- 9 In order to do that one can introduce a

leptonic G. IoU. mechanism in the model. Q,j
Then ~ Q1.j =0
an d the next
J . .
lead~ng term ~s
r~ . . .,r -- if 0(
-
2.~
,~
r
/,~e"~
(~)S.
(~l (<- 'I,..

(4)
M4&.jt )~
!I'j
~'" J J ...... t..
... ,
So (2), (3), ( 4') are three possibilities within the frame-

work of gauge theories. A prejudice of most gauge theorists is
that r~ .. yt behaves as ( 4').

5
In this case ~ ~)rY/~
? D ~~~~'(.
would be of the order of 0( ( ; . ) and then it might be hopeless
for experiments. NeverthelesJ; (1.) and (2) cannot be ruled out
and experimentalists are invited to search for these processes.
Details of the computations can be found in W. Marciano and
A. Sanda (Phys. Lett.67B, 303, 1977).
QUARK-GEOMETRODYNM1ICS: A NEW APPROACH TO HADRONS AND THEIR
INTERACTIONS
G. Preparata
CERN, Geneva, Switzerland and
Universita di Bari, Italy
1. THE EMERGING PICTURE OF STRONG INTERACTIONS
There remains little doubt today that the quark idea must be
taken as the starting point of any serious attempt to build a theory
of hadrons and their interactions. But which direction one should
take appears at the present level of knowledge highly uncertain;
even though the successes of the gauge theories of weak interactions
have given a strong impulse to the belief that the "best" candidate
for a theory of strong interactions is Quantum Chromo-Dynamics (QCD) ,
which has the virtue of extending the gauge principle to the realm
of hadrons. Direct quantitative tests will decide which of the
proposed approaches, if any, should survive; for the time being it
is necessary to stress the importance that all possible avenues be
tried and confronted with the evergrowing experimental information
on the most diverse hadrodynamical aspects.
Be as it may, all attempts to understand hadrons seem to agree

that the following are essential notions:
i) Extended dynamical systems
Hadron extension in space-time is a property suggested by basic

facts of particle physics, like Regge trajectories, form factors
727
728 G. PREPARATA
and the structure of high-energy scattering. All these facts dif-

ferentiate hadrons from "point-like" objects like leptons in a most
far-reaching fashion. On the other hand, it is a feature of known
extended systems (like nuclei) that they are arranged in Regge tra-
jectories, have form factors fastly decreasing in the momentum
transfer, and "diffractively" scatter at high energy.
ii) Quarks
From building blocks of hadronic symmetries [SU(2), SU(3), now
SU(4), and maybe SU(6) tomorrow] quarks have become important tools
first to classify hadronic states in the naive quark model l ) and
after as the elementary scatterers of deep inelastic phenomena 2 ) .
Thus the dynamics of the strong interactions from its low-energy
aspects (spectrum) to the highly inelastic phenomena (Bjorken
scaling) finds a powerful unification through the notion of point-
like s~in ~ constituents whose internal symmetry properties are the
same as the originally proposed quarks.
iii) Colour 3 )
The idea that quarks, besides flavour [SU(3), SU(4) .. ], are

endowed with colour [SU(3) I ] helps us in correlating at least
co our
two hadrodynamical puzzles. The first puzzle is why all known
hadronic states are made out of qq and qqq configurations and their
combinations (triality puzzle), the second is why the wave function
of the lowest baryons (N, ~, .. ) is symmetric in quark spin and
isospin and not, as suggested by Pauli principle, antisymmetric.
The two puzzles can be resolved at once by assuming the existence
of an exact SU(3) colour symmetry and that observable hadrons are
singlets under this symmetry; it then follows that all singlets
must have the configurations qq and qqq and all their combinations,
and that the qqq colour singlet is totally antisymmetric. There
are other puzzles like the value of R = [o(e+e- + hadrons)]1
[o(e+e- + ~+~-)J and the TIo + yy decay amplitude which could find
a solution in colour, but the situation here is much less clear and
more subject to criticism.
QUARK-GEOMETRODYNAMICS 729
It is easy to understand why an approach like QeD lS more and

more looked upon as the only candidate for a theory of strong inter-
actions. It combines the theoretical paradigm of our times, i.e.
Quantum Field Theory (QFT) , with two of the pillars of hadrodynamics:
quarks and colour, and it does this in an extremely elegant and
appealing way, which discloses the possibility of a future "grand-
unification" of all the basic interactions.
If this is so, what is then the motivation to look for some

alternative theory? The reason, as far as I can see it, lies in
the fact that the pillar (i), i.e. extension, is not naturally em-
bodied in QeD, and is related to the problem of "quark confinement"
which is recognized to be an extremely difficult problem 4 ), whose
solution is nowhere near in sight. It may well be that nature is
like that, and that we have to climb the hard way the high wall of
quark confinement. But then, what about the world of hadrons, which
seems so beautiful in its simplicity, and where the results of dif-
ficult experiments can be predicted by playing with the simple toys
of the quark parton model? Is it really true that this wall has to
be frontally assaulted and not circumvented by changing our ideas
about unobservable (confined) quark fields?
It is the necessity to make theoretical contact with the great

experimental achievements of the last years which motivates some of
us to pursue approaches alternative to QeD. These "bag" approaches
start where QeD has so far badly failed, i.e. from point (i). They
are developed by putting "confinement" in from the beginning as a
primitive notion. It cannot be excluded a priori that some of them
will in fact be related to physical QeD solutions, even though such
a possibility seems at present quite remote.
In the "bag" models one assumes that hadronic matter (i.e.

quarks and gluons) can be found only in well-defined finite space-
time regions (bags) and that making these regions unbounded (i.e.
seeing free quarks and gluons) requires an infinite amount of energy,
implying that hadronic constituents are permanently confined.
730 G. PREPARATA
The approaches which have been proposed so far are essentially

of two types:
the MIT-Budapest bagS),
the geometrodynamical theory of quarks and hadrons 6 ).
The basic aspects of the }UT-Budapest approach have been pre-

sented at this school two years ago by Prof. Weisskopf 7 ), while the
first attempts and developments of the geometrodynamical ideas have
formed the arguments of my Erice lectures in 1975 8 ) and 1976 9 ). In
these lectures I shall report on the further progresses of the geo-
metrodynamical theory and shall endeavour to show that it has
reached a rather mature stage in which one can quantitatively attack
several hadrodynamical problems of profound interest.
As already emphasized the starting point of the two descrip-

tions is the same, but their basic strategy is quite opposite. In
the MIT-Budapest bag the geometry of bags originates from a QeD
Lagrangian to which vacuum pressure (and surface tension) terms are
added; in the geometrodynamical bag the situation is reversed, it
is the simple bag geometry which determines the dynamics of the
hadrons (see Fig. 1).
MIT - Budapest Geometrodynamics

DYNAMICS
+
GEOMETRY
Q CD + vacuum and surface tension Simple space-time domains
DYNAMICS
GEOMETRY
Spectrum and perturbative
Bags full of quarks and glue bag interactions
Fig. 1 The basic difference between the HIT-Budapest

and the geometrodynamical bags
2. THE GEOHETRODYNAHICAL PRINCIPLES
Geometrodynamics is not a new word in physics; it was coined

by Wheeler to describe his own extension of Einstein's General
RelativitylO). It correctly stresses the prominent role that the
geometry of space-time plays in determining the dynamical behaviour
of matter, and embodies that momentous train of thought which from
Riemann to Hertz ll ) led to the formulation of the theory of General
Relativity. It is the primary role that geometry should play in
the quark hadron microcosm (bag) that the word geometrodynamics ~n
tends to stress; therefore it should not be confused with some ex-

tension of the theory of gravitation to the elementary particle
realm.
The basic idea of this approach is that the geometry of hadrons

is not the final product of a sophisticated and very involved chain
of dynamical interplays, but that it is a fundamental notion out of
which dynamics follows in the simplest fashion. Quarks are thus to
be viewed as hadronic coordinates describing not only the external
hadron geometry (the bag) but also the internal (the symmetries).
Besides the three notions of extension, quarks, and colour, this
theory rests on two basic principles.
2.1 Simplicity
The fundamental hadrons are quantum mechanical systems, whose

structure is described by the minimum number of quark coordinates.
All hadronic states can be obtained from the fundamental hadrons
by the usual Fok-space construction (+ final state interactions).
2.2 Freedom
Inside the hadronic space-time domains (bags) the quark co-

ordinates carry a wave motion which is the closest to a free motion.
According to (a) and the colour singlet postulate the funda-

mental hadrons are:
732 G. PREPARATA
MESONS, i.e. qq systems,
BARYONS, i.e. qqq systems.
The "derived" hadronic states fall into two classes:
HADRONIC MOLECULES, i.e. nuclei, NN-BB states, charm-anticharm

states,
MULTIPARTICLE STATES, i.e. Fok-constructed from meson and

baryons.
In order to build the theory we must consider three steps:
i) Characterize the fundamental states from principles.
ii) Introduce currents and normalize the states.
iii) Introduce couplings among hadrons.
After this has been accomplished, the description of the dynamics

of hadrons becomes a matter of feasible calculations.
3. THE MESONIC STATES AND THEIR SPECTRill1
According to our principles, we can picture a general rnesonic

state as in Fig. 2. It is characterized by a four-momentum p, a
bounded space-time region (bag) R4(p) in the relative quark coor-
dinate x = xl - X2 , and a set of indices; Dirac (a,S), colour, and
flavour (a,b). Thus to describe such a state we can introduce a
(translation invariant) wave function
(3.1)
which, from the geometrodynamical principles just discussed, can

be characterized as follows:
Fig. 2 The geometrodynamical picture of a meson
i) Confinement
He require that
(3.2)
where R4(p) is a relativistically invariant four-dimensional region,

whose boundary can in general be represented as
(3.3)
ii) Continuity
The wave function (3.1) can be generally decomposed into Dirac

covariants
'P!: l~p<,)= "L pt"(\>J_i})~ tp~~\fi"X.):.

"<" ;II. ...
(3.4)
We require that the scalar functions (r)(p,x)b be continuous func-

a
tions, i.e.
(3.5)
734 G. PREPARATA
iii) Wave motion
The principles of simplicity and freedom lead us to postulate

the following wave equation for the wave function:
(3.6)
where D. = [-i~. + m.JS is the Dirac operator acting on the ith

1. 1. 1. a
quark, and makes it possible to give a precise meaning to the notion
of mass of the quarks.
iv) Maximum freedom
The principle of freedom has not been ex~loited fully 1.n the
wave equation (3.6). Denoting by W(O) the solution of the free
equations
(3.7)
which fulfills the maX1.1nUm possible number of continuity conditions

(3.5), we further require that only those solutions of (3.6) be
retained for which the norm"')
II +_
'\'\0) \\
(3.8)
1.S minimum.
'{e now briefly discuss the physical meaning of the four con-
ditions which characterize the discrete "fundamental" meson states.
*) After the "conventional" normalizations 111jJ II = 111jJ(O)11 1 has

been imposed. The norm can be defined as
or as
Max T<-"~~) .
)I.E: 1Z'c.~)
QUAR KGEOM ETRODYNAM ICS 735
Condition (i) requires that the quark degree of freedom, or better

coordinate, be confined in finite space-time regions (bags) whose
structures can be characterized by the two functions Rs (0)2
. and
R (p)2 appearing in (3.3). Note that no complicated "bag" surface
t
configuration can appear in this formulation, because here, dif-
ferently from other "bag" approaches, the dynamics of the meson
systems is totally described by their quark coordinates. He shall
discuss later the important criteria which constrain the p2 depen-
dence of Rs and Rt
Condition (ii) allows us to define a finite quark (antiquark)

momentum operator
!,.,.
2
-'
In fact were (3.5) not satisfied the momentum operators applied to
tpe (r)'s would yield infinity at the boundaries. Note that (3.5)
is the strongest continuity condition which can be imposed on the
wave function compatible with the wave equations (3.6).
Condition (iii) specifies the differential equations obeyed by

the wave function when x is inside the "bag" region R4 (p). The
choice (3.6) is suggested by the simplicity postulate and the idea,
central to this approach, that geometry [i.e. the structure of
R4(p)] plays the fundamental role. Inside R4(p) the "quark motion"
is the simplest one compatible with the boundary conditions (3.5).
It is perhaps worth recalling that (3.6) is the form to which a
Bethe-Salpeter equation for two spin-i objects reduces when the ker-
nel (or potential) vanishes.
As for (iv) it will playa crucial role in determining the

structure of the solutions of (3.6); it deserves therefore some
more extended comments. Equations (3.7) are not only the simplest
set of equations which can be written for a meson wave function,
but also embody an aspect of quark dynamics which ~s of paramount
importance, i.e. the free behaviour of quarks inside hadrons. The
736 G. PREPARATA
puzzling successes of the simple parton models testify to the phy-

sical relevance of free quark behaviour. Equations analogous to
(3.7) were initially proposed 6 ) in the context of scalar and iso-
scalar quarks; however, it was noticed that the continuity require-
ment could not be implemented on the time boundary [(pX)2/p2 =
= Rt (p2)2J owing to the first-order character of the relative time

equation. The idea was then to construct equations for the meson
wave functions, which generalized (3.7) while allowing compatibility
with (3.5). Equation (3.6) is one such generalization, but it has
the property of admitting a class of solutions which have no counter-
part in the class of "free solutions" ljJ(O) and seem physically un-
acceptable. The condition of approximate freedom thus gives a pre-
cise meaning to a solution of the simple equation (3.6) which ~s
continuous [Eq. (3.7)J, and resembles as much as nossible the free

wave function ljJ(O).
He shall now proceed to exploit the four previous conditions

to derive the meson spectrum and the corresponding wave functions.
We begin by analysing the solutions ljJ(O) of (3.7). The procedure
we adopt is to develop ljJ(O) in a Dirac basis:
(3.9)
where r R are the 16 Dirac matrices [R = S, P, V, A, TJ. By pro-

jecting on the various Dirac covariants r R , we obtain 10 equations
which the ~R's have to satisfy. A straightforward analysis gives
us two families of solutions:
i) the pseudoscalar family
'P")! = L\'.(rtM;)"'s (flTMz. Y)! 4>~") \~; ~) (3.10)
where p. -1/i(3/3x )y~, and ~~O) obeys the two equations

~ ~
(3.11)
K. being the Klein-Gordon operators (0. + m~).

1 1 1
ii) The vector family
(3.12)
where ~(O), in addition to obeying Eqs. (3.11) satisfies the trans-

).l
versality condition:
(3.13)
The solutions of the system (3.11) have been discussed at length in

Refs. 6, 8, and 9. He only recall that in the rest frame
[p
).l
= (M,O)] a function ~(O)(p;x) satisfying (3.11) has the following
structure (we set for simplicity m1 = mz):
(3.14 )
with
(3.15 )
and
(3.16 )
The solutions of (3.15) which vanish at the space-boundary

I-+xl = Rs (M) are described by two quantum numbers, angular momen-
y:
tum, n radial quantum number, and are given by
cPnt (,~) =Nnt l..( (" ~nt \~I ) (Q~) (3.17)
where Nn is a normalization factor, j(x) the spherical Bessel

function of order , and
738 G.PREPARATA
~nf -~
- ~lt\l") ) (3.18)
th
Bn being the n positive zero of j(x). From (3.15) it also fol-
lows that
(3.19)
which gives us a discrete spectrum in terms of the quantum numbers

nand . As for ~M(t) the only non-trivial solution is
""fill It) _ urn$\: ~

which cannot meet the continuity requirement at the time-boundary
Summarizing, the solutions of (3.7) can be grouped in two

families, the lowest energy members of which are a 0- and a 1-
state, respectively. Their spectrum is labelled by the two quantum
numbers and n according to (3.19), once R (~1) has been given the
s
corresponding energy levels are predicted. The physical relevance
of these solutions for the observed meson spectrum needs no further
comment.
We are now in the position of finding which are the mesonic

states. The results just derived for the ~(O)'s lead us immediately
to look for those solutions of (3.6) which can be cast in the forms
(3.10) and (3.12). Thus we obtain:
i) The pseudoscalar family, which is given by
(3.20)
where ~5' according to (3.6), obeys the equation
(3.21)
ii) The vector family,
V! = [(y'+Yh.')'ifllPf+M J! ct>"'lf~X)
L )
(3.22)
where , in addition to obeying (3.21), satisfies the transversality

~
condition
(3.23)
We study now the solutions of (3.21) which, ~n the sense of (iv)

above, are as close as possible to the solutions of the system (3.7).
Taking
(' Z. 2
Q.nd o:ml-m to ,
we can write (3.21) in the rest frame
(3.24)
-+ -+
Decomposing (x,t) M(x)I/!H(t), as indicated by (3.14), (3.24) be-
comes
(3.25)
and
[(tl +~_A_M")L_ lMto- i)"] 4>", lb) = 0 J (3.26)
4 2
with the boundary condition [see (3.5)J

4>M (lx\ =lZs) = 0
~" Ob\ = ~t ") = 0 .
The solution of (3.25) is immediate, and coincides with (3.17) if
we set
2.
2 ~r'le
A= "".(=-& (3.27)
12$
The general solution of (3.26) is of the form
(3.28)
where w. are the four solutions of the quartic equation

~
(3.29)
740 G. PREPARATA
We get
(3.30)
where we have introduced the "quark energies" E.

~
/iFnQ, + 2
m ..
~
A~proximate freedom eliminates immediately w~+) and w[-),

which give rise to ranid oscillations in W~1(t). The condition
WM(R t ) = 0 yields
,lob c P-' ic) ~

~"lt:"): N [e 1-+ (:-i) e. 2] (3.31)
where
CUI = - "" + E1
2-
(3.32)
~2. = ~ - E:z
2
and the eigenvalue condition (p integer positive)
(3.33)
Approximate freedom again comes to our help by excluding all

the solutions with p > 1. We finally get
(3.34)
and combining (3.32) and (3.33) we can write the very simple eigen-
value equation for the mass of the state (nQ,):
(3.35)
where
(3.36)
and knQ, ~s given ~n (3.18).

The physical interpretation of (3.35) is particularly sugges-

tive; it tells us that the mass of the meson state is Riven by the
sum of the quantized energies of its two quark waves m~nus a con-
tribution, stemming from the finite extension of the time bag, which
resembles a binding energy term.
-+
The extens~on of M(x,t) to moving frames is straightforward
and has been discussed in last year's course 9 ).
Equations (3.35) and (3.36) just derived predict the spectrum

for the two meson families, the pseudoscalar and the vector, once
R (H2) and R (H2) are specified. I t is clear that through such
s t
equations Rs and Rt are functions of the geometrical constants Bn
only. The Massive Quark ~1odel criterion of asymptotic universality
for the distributions of "quark masses", discussed at length in
previous courses 8 ,9) gives the asymptotic behaviour
Rs 12~ ~ R,2).\. (3.37)

) l'\~~
The simplest parametrizations of R (S n) which, via (2.27) and

- s, t nX,
(2.28), are compatible with (3.1) are
'& 2
'R, = 'Ro ~"( (3.38)
~:= ~:~t'" ~ (3.39)
with R~ = 2R 2 , and C constant.
It should be noted that inserting (3.37) and (3.38) in the

spectrum equations (3.35) and (3.36) yields the unperturbed energy
levels. The unitarity corrections coming from self-energy diagrams
such as
+ +
742 G.PREPARATA
which must be considered once we introduce the three-bag counling

(see later), will be responsible for mass shifts as well as for
finite widths of the meson states. Thus a precise determination of
the parameters entering in (3.38) and (3.39) must await the com-
pletion of the fairly involved program of computing such effects.
The previous results should be looked at having this always in mind.
3.1 The Pseudoscalar Nonet
This very important subfamily deserves special discussion. So

far our outlook has been typical of a quark model approach, in which
the mass differences between the lsi = 0 and the lsi = 1 states
should be attributed to fine structure effects such as the just
mentioned unitarity corrections.
However, there is another aspect of pseudoscalar dynamics which

must be taken into account, i.e. the chiral structure of the hadro-
nic world. We shall check later that when m ~ 0 the axial currents
q
become conserved [with the possible exception of the ninth*)] and ~n
order to avoid parity doubling the pseudoscalar octet must choose
the Nambu-Goldstone mode. Thus according to (3.35) in the limit
m ~ 0, w ~
q
21tl for the eight pseudoscalar mesons (n,K,n) while
all other states are expected to be relatively unaffected by small
variations in the quark masses.
In order to implement w ~ 21tl ~n our approach we have three

choices: to change Rs in (3.38), to modify Rt in (3.39), or to
change them both from the value they assume ~n the vector meson
subfamily. It seems natural, however, to maintain the quark model
outlook (n-p degeneracy) in the space dynamics and to modify time
dynamics, i.e. to let Rt vary. For it is the time coordinate that,
being conjugate to masses, responds to the symmetry-breaking effects.
Needless to say, understanding the mechanism of chiral as well as
*) In order to overcome the famous U(l) problem one must assume that
there is some additional piece to the ninth axial current 12 )
SU(3) breaking in the context presented here is of the utmost im-

portance, but so far such an insight is unfortunately lacking.
Accordingly, Eqs. (3.35) and (3.36) assume the role of deter-

mining Rt of the wave functions of the pseudoscalar octet, once Rs
has been fixed at the same value as the vector nonet. In the case
of the n-Xo system, SU(3) breaking requires the introduction of two
wls according to the equations
(3.40)
(3.41)
where, as usual, E
p,A
= ~2 + m2 ,.
P,/\
The requirement that the masses of the nseudoscalar octet de-
pend linearly on the chiral symmetry-breaking parameters (mp,mA),
as we shall see later on, through the s~ectrum equations, deter-
m~nes the low-mass behaviour of w(t!) as
(3.42)
where wo = 2\k 10 \' This has the consequence that the pseudoscalar
octet obeys a quadratic Gell-Hann-Okubo relation; ~n fact to first
order in m2 and m~:
p /\
(3.43)
2 'a.
Mp+mJ\.
(3.44 )
'2. '!c'o ,
z 'Z
""p ... 2",,,-
(3.45 )
6 '''10\
where the last equation follows from (3.42) by setting, as required
by the n-wave function,
lp) 2 l.\)
00,,= !OO'l~3~'l ,
where w(p) and w(A) are defined in (3.40).
n n
744 G.PREPARATA
3.2 The Other ~1eson Nonets, before Fine Structure Effects
All other meson states have no special role to play with re-
gard to the patterns of hadronic symmetries and their spectrum must
therefore follow the simple structures of Eqs. (3.35) and (3.36).
lve predict that all meson families are, before unitarity cor-
rections, in the ideal mixing configurations with the I = 0 non-
strange meson degenerate with the I = 1 member of the octet. In
Table 1 the calculated meson masses are reported with the following
parametrizations in Eqs. (3.38) and (3.39):
~: =4 G~\l2. (3.46 )
c =-3 ~v2. (3.47)
mf= O.Ol Ge.\I'Z (3.48)

....
m~= 0.2'2. Ge\l-z. (3.49)
Table 1
The calculated spectrum for SU(3) states
~ 0 1 2 3 4 5 6 7 Particle type
0.76 1.32 1. 71 2.02 2.29 2.53 2.75 2.94 p-w

1 0.88 1.41 1. 79 2.10 2.37 2.60 2.81 3.01 K*
1.00 1.51 1.88 2.18 2.44 2.67 2.88 3.07 q,
1.85 2.19 2.48 2.73 2.95 3.16 3.35 3.52 p-w

2 1.93 2.27 2.55 2.79 3.01 3.22 3.40 3.58 K*
2.02 2.34 2.62 2.86 3.08 3.28 3.46 3.63 cjl
2.54 2.81 3.05 3.27 3.47 3.65 3.82 3.99 p-w

3 2.61 2.88 3.12 3.33 3.52 3.70 3.87 4.04 K*
2.68 2.94 3.18 3.39 3.58 3.76 3.93 4.09 cjl
3.09 3.32 3.53 3.73 3.9l 4.08 4.24 4.39 p-w

4 3.15 3.38 3.59 3.78 3.96 4.13 4.28 4.43 K*
3.21 3.44 3.64 3.83 4.01 4.17 4.33 4.48 cjl
3.56 3.77 3.96 4.13 4.30 4.46 4.61 p-w

5 3.62 3.82 4.01 4.18 4.35 4.50 4.65 K*
3.67 3.87 4.06 4.23 I 4.39 4.55 4.89 cjl
3.98 4.16 4.34 4.50 4.56 p-w

6 4.03 4.21 4.38 4.55 4.70 K*
4.08 4.26 4.43 4.59 4.74 cjl
4.35 4.52 p-w

7 4.40 4.57 K*
4.45 4.61
I cjl
QUAR K-GEOM ETRODYNAM ICS 745
One obtains a set of approximately linear and parallel Regge tra-

jectories without odd daughters. The agreement on the highest J
members of the ,n families is remarkable. The SU(3) symmetry
pattern comes out in very good agreement with experimental informa-
tion. In Fig. 3 we give the enerp,y levels in a spectroscopic nota-
tion; one should note the mass bands making up distinct fire-
sausage structures as we shall discuss later. No attempt has
5 s p o F G
6- 5- Firesausage band
7- 6-- 5-
7- 6- 5-- 4-- M= 4GeV
6-- 5-- 4---
4~6~~~~5~~~-'4~~~~----~~-
5 4 5-
5--- 4- 3-
4-- 3-- 2--
2---
3 3-- 2-
3-- 2-- 1-
3- 2-- 1-
1-
2-
2
2- 1--
1--
1--
1-
Fig. 3 The spectroscopic scheme for the

meson states of the p-w type
746 G. PREPARATA
been made to optimize the values of the four parameters (3.46) to

(3.49), because such an operation would assume a precise physical
meaning only after the first-order unitarity corrections have been
properly calculated. In view of this, the lack of LoS splitting
exhibited by the calculated spectrum cannot be taken at this stage
as a legitimate cause for worry.
3.3 Hidden and Apparent-Charmed States
In order to calculate the spectrum for both the charmed states

and the members of the J/W family we need only to introduce a value
of the charmed quark mass m . The pseudoscalar mesons are expected
c
to show some trace of an otherwise badly broken chiral symmetry and
our inability to grasp this problem prevents us from making definite
predictions. We expect, however, the chiral symmetry-breaking
effects to be much less severe on this state and that the vector-
pseudoscalar splitting can be correctly given by the unitarity
effects.
The calculated spectrum 1S presented 1n Table 2 for the various

(,n) by taking
(3.50)
It should be noted that the mass of the D* is 1.93 GeV against the
experimental value ~ 2.0 GeV; but again shifts of this order of
magnitude are no cause for worry. Of particular interest is the
closeness of the vector states belonging to the subfamilies [(,n)]
(O,n) and (2,n-l). Such states as an effect of unitarity will mix
quite thoroughly, thus giving rise to mixtures of S- and D-waves.
Thus we predict the W'(3.7) and the W"(4.02) to be almost per-

fect mixtures of the (S,2) and (D,l) states and both have a sub-
stantial coupling (see Section 7) to the photon, in spite of the
fact that the D-wave state coupling to the photon is quite suppressed.
Table 2
The calculated spectrum for charmed states.
~ 0 1 2 3 4 5 Particle type
1.93 2.39 2.71 2.97 3.20 3.40 D

1 2.03 2.48 2.79 3.04 3.27 3.46 F
3.10 3.47 3.72 3.93 4.11 4.28 ljJ
2.83 3.12 3.36 3.57 3.76 3.94 D

2 2.90 3.18 3.42 3.63 3.82 3.99 F
3.82 4.04 4.24 4.41 4.57 4.72 ljJ
3.41 3.64 3.84 4.04 4.21 4.37 D

3 3.47 3.70 3.90 4.09 4.26 4.42 F
4.28 4.48 4.65 4.81 4.95 5.09 ljJ
3.88 4.08 4.27 4.44 4.60 4.75 D

4 3.94 4.14 4.32 4.48 4.64 4.79 F
4.67 4.85 5.00 5.15 5.29 5.42 ljJ
4.29 4.47 4.64 4.80 4.85 5.09 D

5 4.34 4.52 4.68 4.84 4.99 5.13 F
5.02 5.18 5.33 5.47 5.60 5.72 ljJ
4.66 4.83 4.98 5.13 5.27 D

6 4.70 4.87 5.02 5.17 5.31 F
5.34 5.49 5.63 5.76 5.89 ljJ
A similar situation obtains for the states (S,3) and (D,2), which
we predict at 4.28 and 4.24 GeV, respectively. They can account
for the bumps at 4.10 and 4.4 observed at SPEAR. Finally the mass
differences between D and F states come out around 100 MeV.
What happens if there exists another heavy quark with mass
(3.51)
'YTl..a = S.l GftV ?
In Table 3 we can read predictions which may be relevant for the

spectroscopy of the T's.
748 G. PREPARATA
Table 3
The calculated spectrum of the T family (mu 5.2 GeV)
~ 0
9.53
1
9.80
2
9.97
3
10.10
4
10.21
Particle
T (H + H)
ty~e
1 5.15 5.56 5.84 6.06 6.25 :J) (H + p)

5.26 5.65 5.92 6.13 6.32 :7 (H + Ie)
10.03 10.17 10.28 10.38 10.47 T
2 5.94 6.lS 6.38 6.56 6.71 ~
6.02 6.25 6.45 6.62 6.77 S-
10.31 10.42 10.51 10.60 10.68 T

3 6.43 6.62 6.78 6.94 7.07 ;J)
6.49 6.68 6.84 6.99 7.13 :7
10.53 10.62 10.71 10.79 10.S7 T
4 6.81 6.97 7.12 7.26 7.39 ~
6.87 7.03 7.17 7.31 7.44 :7
10.72 10.81 10.S9 10.97 11. 04 T
5 7.14 7.29 7.42 7.55 7.67 :J)
7.19 7.34 7.47 7.60 7.71 S-
10.90 10.98 11.06 11.l3 11. 21 T

6 7.44 7.57 7.70 7.S2 7.93 :J)
7.48 7.62 7.74 7.S6 7.97 :7
4. BARYONS AND THEIR SPECTRUH
The application of the geometrodynamica1 ideas to the qqq

system has been worked out 1n collaboration with Szego 13 ). The
baryon wave function shall be written as
(4.1)
where a, S, yare Dirac indices, a, b, c carry flavour and colour,

and
QUARKGEOMETRODYNAMICS 749
(4.2)
The characterization of (4.1) proceeds exactly as ~n the meson case:
i) Con.Nnement
(4.3)
:>
where R8 (p;x,y) is a compact eight-dimensional space-time region

with boundary B8 (p;x,y).
ii) Continuity
Decomposing into Lorentz covariants (4.1), ~.e.
,1\ ""f\ 'LJ~) .'))~)) (4 4)

"alCl~blC \P~~~)= ~ L-<P"C' l~-,-'~JI;I:.1 CPabctp~)t/' .
(r)
we require that the scalar functions abc(P;x,y) be continuous func-
tions, i.e.
(4.5)
iii) Wave motion
For x,y E R8 (p,x,y) ~ obeys the simple differential equation
(4.6)
where D. is again the Dirac operator.

~
iv) Maximum freedom
The "distance" of ~ from the "free solution" ~~~~i3cy(p;X'Y)

determined by the free equations
750 G. PREPARATA
(4.7)
1S minimum, 1n the sense of (3.8).
In solving the ~roblem for W(O) we get a big surprise. In

fact we can find no solution of (4.7) in a bounded space region,
unless the space coordinates (in the rest frame) are constrained by
the relations:
(4.8)
Such a constraint corresponds to a spatial configuration where the

-+
three quarks are aligned along the direction of the vector R. This
constitutes an important difference between the geometrical and the
field theoretical approaches to the quark degree of freedom, it
being very hard in the latter to understand how such a constrained
motion could arise. Choosing ~o = we obtain for the spatial
structure of the baryon a quark-diquark confieuration which seems
to be favoured by present experimental information.
Proceeding now in a way completely analo8ous to the meson case

we obtain the following results.
a) For a completely symmetric baryonic wave function the physical

states can be organized in [SU(6),L P] multiplets following the ~at
tern of Fig. 4. Note that the representations (20,L P ) and (70,0+)
L
are absent, and that the ~arity of the states 1S always (-1) .
These are all consequences of the suppression of three degrees of
freedom dictated by our a~proach, and seem to have experimental
support.
b) The mass of a physical baryon state is given by the equation
(4.9)
-~
MGeV
lP: 0- ,- 2- :r 4- S-
..2Q..
---1Q..
2i.22..
....2Q... -2L.2!L 22-
..2Q.. 22-
2L ..li..2!L 22-
..2Q.. A.2!L
..2Q..
-2L .iL2Q.. ..2Q..
2..22..
3 22- 22-
.iL .....2Q.. 22-
~2Q..
..1Q...
..:L
2 -2L -li.1Q..
..1Q...
~
1
Fig. 4 [SU(6),LP] pattern of physical baryon states
where W[SU(8)](abc = 1, ... , 4; i, j, k = 1, 2) are the normalized

SU(8) wave functions (m. are the quark masses);
~
(4.10)
(4.11)
00::. 1t. (4.12)

~t I
nl. }
where Rs and Rt have the same meaning as in the preceding paraeraph,
and can be parametrized similarly to (3.38) and (3.39). See Table 4.
752 G. PREPARATA
Table 4
The mean masses of the [SU(6),L P] multiplets
~ 1 2 3 4 5
(56, 0+) 1.14 2.07 2.71 3.24 3.69
(70, 1-) 1.60 2.39 2.97 3.46 3.89
(56, 2+) 3.20

1. 94 2.65 3.66 4.07
(70, 2+)
(70, 3-) 2.23 2.89 3.41 3.85 4.24
(56, 4+) 2.48 3.10 3.60 4.03 4.41

(70, 4+)
(70, 5-) 2.70 3.30 3.78 4.19 4.56
(56, 6+)
2.91 3.48 3.94 4.34 4.71
(70, 6+)
(70, r) 3.10 3.65 4.10 4.49
(56, 8+)
3.27 3.81 4.25 4.64
(70, 8+)
5. THE INTERACTIONS OF HESONS
In this section we shall construct the interactions among

mesons starting from their geometry. The general strategy of how to
compute hadrodynamics starting with the fundamental hadron inter-
actions has been described at length in the Erice courses of the
last two years and need not be repeated here.
He shall begin by exhibiting explicitly the structure of the

meson wave functions, which shall enter in an essential way ~n the
construction of meson interactions.
5.1 Meson Have Functions
In Section 3 we have seen that as a result of the general

characterization of meson states their wave functions are given by
(5.1)
(5.2)
for the pseudoscalar and the vector families, respectively. Both

s(p;x) and W(p;x) obey the equation
~1\Ca. ~ lpj'X.) =0 (5.3)
and we can write
4>s lp~x) = Q> (pjX) (5.4)
4>,..('; ')I.) =;.<.p) ~ ~Pi ').) J

(5.5)
where (p;x) is the solution of (4.3) which is "closest" to the free

solution (see Section 2), and w(p) is a polarization vector obeying
pW (p) = O.
W
According to (3.14) in the rest frame p = (11,0),
-+
we have
(5.6)
where
<:PM fttlt) = \4~'!!'~ )~1.2-- 3 (Il,,! \1) ) ...,,"'\0.. )

1t 12
'.1" 4' (0 )
t ...~ ~n.
A. r n
~'I n e
.(. )1;,
which has been conventionally normalized in such a way that
and
754 G. PREPARATA
(5.7)
with
(5.8)
ti)- H
~-
2
which has also been conventionally normalized such that
The quantity wn defined as
(5.9)
is related to R n by the equation

t,n",
1t"
.i)nl = (5.10)
~,ne
The wave function (p;x), in a general Lorentz frame in which
the meson is moving, can be obtained by (5.6), (5.7), and (5.8) by
the following identifications:
(5.11)
(p)(,) (5.12)
"1i\
and ~ 1S given by the direction of the spatial part of the four-
vector obtained from x by applying the Lorentz transformation A-1(p)
which brings the meson to its rest frame; i.e.
(5.13)
He shall now calculate the Fourier transform <!>(p;k) of <!>(p;x):
As a result of a simple computation, we have in the rest

frame:
(, .. ) (4ft )I;z. ~ 2 (L \Wt (.1zs,n' i~') .

4>M ntI
\.k:o ~
)
-. -
Iii
"5 Tt '"'TIl'
(~, itl)Z_ fl:t
(5.15)
CII& [2~.nf tko-+ 7- )J

2
~1. _ c.~o . . CI), y' )
where all quantities have been previously defined.
The extension to a general Lorentz frame is again immediate,

and consists in identifying in (5.15)
(5.16)
(5.17)
and
(5.18)
A convenient approximation of <!>n(p;k) ~s
(5.19)
where 0R 2 (z) (R 2 /TI) (sin z/z) is the "fat o-function" introduced ~n

Ref. 2.
756 G.PREPARATA
5.2 The Three-lieson Coupling
The general structure of meson couplings has been discussed

~n last year's lectures 9 ). There it was shown that, in order to
accommodate all the salient dynamical properties of mesons, only
the three- and the four-meson couplings are needed. Here we shall
give a general description of the three-meson coupling only; four-
meson couplings have not been worked out in detail yet.
The diagram describing the three-meson coupling ~s depicted

~n Fig. 5, where the dots represent a two-body "quark operator"
denoting the probability amplitude for a quark of one meson to tun-
nel into another meson in the space-time region where the two mesons
overlap. The most general form for such an operator ~s
(5.20)
++
where p = (l/2i) ca/'dx ). The actual form of the functions A and B
U ]J
is important only for the low-lying mesons; for high masses ~n
fact the operator p2 has the universal eigenvalue Mw = 'IT /R2.
Several choices of A(p2) and B(p2) are, however, possible and

~n the calculations to follow the particular choice was made that
(5.21)
P,
P3
Fig. 5 The three-meson coupling
which corresponds to an inverse propagator of a quark which cannot

propagate through the vacuum, and ~2 is a constant which has the
dimensions [massJ2 and plays the role of a coupling constant. Its
value can be determined by fixing one coupling constant, say g ,
pmf
at its experimental value.
The general form of a three-meson vertex after substituting

(5.21) is very complicated and need not be given here; we should,
however, keep in mind that once the wave functions' normalizations
are determined (see later) and a simple parametrization of (5.21)
has been supplied, we are in a position to compute all three-meson
coupling strengths.
The extension of the meson couplings to the unphysical rerion

has already been discussed at this schoo1 9 ), the procedure is
tedious, but otherwise well defined and ~resents no particular
difficulty in its implementation.
6. CURRENTS AND WAVE FUNCTION NORUALIZATIONS
In order to determine the physical normalization of the wave

functions we must first define the action of the current operators
in the space of the meson wave functions and then require that the
charge of hadron states be given correctly. In Ref. 9 we have
analysed the problem in the context of scalar quark coordinates,
and the conclusion was reached that in order to have a consistent
description of high-energy quark-quark scattering it is necessary
that the physical current operators be made up of two distinct
pieces: the vector-meson dominated and the direct one. The dif-
ferent roles of these two contributions, and their important phy-
sical consequences have been adequately stressed 9 ); in this sec-
tion we shall apply these ideas to the meson case, work on baryons
is still in progress.
758 G. PREPARATA
6.1 The Vector Dominated Piece
The vector-meson dominated piece of the vector currents ~s de-

fined through the sum of the diagrans ~n Fig. 6. The vector domi-
nated form factor ~s then given by the expression
(6.1)
where V(V ~ AB) ~s the three-meson coupling discussed in the pre-

n
ceding section, and
= Z
,
N
V"
Sd\
~Jt)""
,.~ ()., y.
2. n
) (". ,,)
'In \: ) .
(6.2)
In (6.2) A. are the Gell-~1ann matrices, Zl is the current norma-

~
lization factor (in a theory with colour quarks Zl = 13)

and NVn ~s
the normalization factor of the vector meson V. By taking A = B
n
in (6.1) we obtain the vector-meson dominated form factor of the
given meson A. Putting A./2
~
= Q, selecting the charge form factor,
and setting q2 0, we get the vector meson contribution to the
V
charge QA' to be called QAZA' in the form
Fig. 6 The vector-current piece dominated by

vector-meson intermediate states
Obviously Z~ is proportional to Nl; thus we can write in general
(6.4)
where WA can be explicitly computed once the normalization problem

has been solved for the vector mesons Vn . The calculation of l1A
is, in general, a quite difficult non-linear problem; a drastic
simplification, however, occurs when we realize that, owing to the
structure of the overlap integral I(V + AA), the contribution
n
which is definitely dominant in (6.3) is the one corresponding to
the vector mesons of lowest mass p, W, ~, and J/. This observation
enables us to set up an iterative calculational scheme for WA
In the following we shall only kee~ the lowest vector mesons (V)
and write approximately
(6.5)
where
(6.6)
6.2 The Direct Piece of the Current Operator
We must now construct a vector current operator which couples

directly to the quark coordinates and Meets the requirements
i) that it is conserved [in the SU(3) limit];
ii) that it is additive in the quark charges.
In Fig. 7 the diagram corresponding to a vector current is

reported where the operational definition of the symbols requires
a more detailed discussion which has been carried out in Ref. 14.
In particular, by taking two equal particles, one obtains at zero
momentum transfer:
760 G. PREPARATA
3
~ 3
= IK '3
~~/2

3
(jf3- m 3)
IK 23
Fi 0. 7 The action of the direct piece of the

current operator Ji(O) on meson states
].1
(6.7)
Writing, as suggested by the structure of the direct current,
(6.8)
the normalization factor NA is then determined by the equation
N"l. 1 (6.9)
A=
WA+ '}.A
In this way we have achieved a well-defined ~rocedure to nor-
malize the meson wave functions. After this is accomplished every
hadronic property can be computed starting from the basic parameters
of our theory [R~, c; mp ' rnA' mc ' ... , ].12J. In the next section
the first preliminary results on meson properties will be reported.
7. THE CURP~NT-PARTICLE MATRIX ELElffiNTS
In this section we shall compute matrix elements of a current

between the vacuum and the one-meson states, and compare the re-
sults with experiments.
7.1 The Coupling between a Vector Current and a Vector State
We consider the matrix element

From (6.2) we readily compute for the = 0 states
where from (5.7) we can write
(7.2)
Asymptotically [MV ~ ooJ we have
For the 2 states we calculate a much smaller coupling, given by
which becomes asymptotically
(7.5)
which is a factor ~ 20 down from (7.3).
The coupling of = 2 states to the nhoton does not vanish,

as in the non-relativistic limit, but it is nevertheless very sup-
pressed with respect to the = 0 states.
Choosing the value Zl = 1:3, as suggested from a colour theory

of quarks (note, however, that in this approach Zl is essentially
762 G. PREPARATA
arbitrary), we can determine Np by requiring that Yp be given by

its experimental value*)
and we obtain
(7.6)
From (7.1) we can also compute Yw and y and obtain
(7.7)
This agrees with experiments; and it is quite remarkable that the

yls follow a SU(3) pattern, and not m2 jy or any other combination
of m2 and y. Incidentally m2 jy deviates from the SU(3) pattern by
some 70%.
We compute a(e+e- ~ hadrons) by considering the sum of the

vector meson contributions of arbitrarily large masses. (See Fig. 8.
In Ref. 9 this problem was studied in the case of scalar quarks;
here we shall proceed in a completely analogous way.
We define the vacuum polarization tensor
"IT,,~ lQ) = \.QIlQ, - ~\L.fl) ltlEt) ::: (7.8)

_6 l11l.)~ ~'t l~-"pn) <,0\ ~lC)) \'I\,)<n 1~ ~t()) \0') J
n
~ 1 JVVVV():;- 12
Fig. 8 The diagrams contributing to n(Q2) [Eq. (7.9)J
*) This requirement is not independent of the value of the coupling

gpon+n- as in standard vector dominance.
where
and (Mf) +constant 1S ) at high energy. In the high mass re~ion we

n
have (for = 0)
(7.10)
and we can easily evaluate the sum (7.9), which p,ives
(7.11)
From the normalization condition neglecting, as it 1S legitimate

being of order 11M, the vector dominated piece, we obtain
(7.12)
Putting all factors together, we finally obtain
2:~.~
I (7.13)
I
I
(7.14)
we obtain the remarkable result
4 Z2. z~:z. (7.15)

1(. 1 .
\
I
Setting Z21 = 3 the constant in front of L.O~ has a value 3.82;

1 '1
thus we calculate
'R.-. 2.SS 3 "u.aC'~s lpJn,"A )
.... 4,2S 4 ""'4o('~ (yJ",A,C,),
764 G. PREPARATA
This also looks quite good; however, one remarks that the effect
of the direct y-coupling as well as the finite mass corrections
have been left out. All these problems will be analysed in due
course.
7.3 The e.m. Hidths of the IN Family
Given the Yv's whose ex~ressions are exhibited in (7.2) for

= 0 states and In (7.3) for = 2 states, the vector meson lep-
tonic widths are computed through the well-known formula
C'("~tt~-) .... Xr/.."2 Mv. (7.16)

3 .... ~
V
Accordingly, we can write for the ratios of e.m. widths
'Z
1"\ "V
= -
M,,'-
. . . .;:1.
-(" '
(7.17)
By use of (7.2) we have for any radial recurrence of the J/W par-
ticle; W(n):
(7.18)
For the J/W we then compute [N W = 0.23J:
r (~/~ ... ott.) rC O.l:.

r(f>~ ~1t-)
To compute the hifher widths we need only to compute NWn'

which gets determined by the direct coupling only (the J/W cou~ling
lS substantially suppressed by an unfavourable overlap). He get
(7.19)
which gives for n = 2:
" IJ"N>~ .e~)

r l'\>l\)... .t"'.c)
1.e. This is not 1n disagree~ent with the
reported ~ 1.8 keV for the ~', for we have argued in Section 3 that
substantial mixing should occur between the = 0 and the = 2
vector state. In the maximal mixing configuration the figure
3.3 keV should be halved and attributed to each of the (3.69) and
(4.02) states. This agrees with observation.
7.4 The Coupling between an Axial Current

and a Pseudoscalar State
The matrix element to be computed is

,. (l
<O\A~ (0)1 pi>

By using our wave function we compute
= Z1
~ b
'P) (op)
(7.20)
with
For the n+ we get:
(7.22)
and for K+
(7.23)
766 G. PREPARATA
Evaluating the various normalization factors following the

procedure outlined In the preceding section, and recalling that
m~
1\
= 0.22 GeV 2 and m2
p
= 0.01 GeV 2 , we obtain
1.1S
He can furthermore compute the absolute value of f +. The result

IT
is
1,.1' - C.B G~.
Experimentally, we have
The agreement IS very good.
We end this section by stressing that several very demanding

tests have been passed by the approach presented here with great
ease. However, whether or not we possess an accurate theoretical
tool to deal with hadrodynamics will be answered only by a much
wider and more refined analysis.
8. GEOHETRODYNAmCS AT HIGH ENERGy 16 )
In this section I will review the results so far obtained In

the description of high-energy hadronic scattering processes. Some
of the important points of the high-energy behaviour in the geometro-
dynamical theory have been touched upon in my previous Erice courses;
in this section I would like to emphasize the coherence and the
interdependence of the many different aspects of high-energy scat-
tering which emerge in this approach.
8.1 The Fire-sausage: The Prototypical

High-Energy Hadronic State
We have seen that for "low-mass quarks" (p, n, A) the spectrum

of fundamental hadrons [mesons (qq) and baryons (qqq)] turns out as
comprising an infinite number of almost linear and parallel Regge

trajectories without "odd daughters".
In this treatment, which is just a first a~proximation, I shall

consider only the meson states; the baryon states in fact turn out
to be less coupled at high energyl?).
The spectrum of such states is depicted in Fie. 9; the line

[RrUl) 'V log MJ
l.mc)~ = (8.1)
defines the boundary between the meson states which have a wave
function of order 1 ( ~ ) and those which have an exponentially
max
o 10 20
M2 GeV
Fig. 9 The meson Regge trajectories in the

(orbital-angular momentum)-M 2 plane
768 G. PREPARATA
small wave function ( > ) 18). Thus at a given high mass M we

max
have a set of ~ /2 states approximately degenerate in mass
max
which couple strongly to their quark-antiquark constituents (large
wave functions). By means of these states we can construct a set
of "coherent states" by taking the following superposition:
(8.2)
where ~ n (p,k) is the wave function describing the meson state

n",m
(n,) whose mass is given asymptotically by
(8.3)
These states have a very simple structure in configuration space.

It is, in fact, very easy to show that the states (8.2) correspond
to a cylindrical space domain (see Fig. 10), which we shall call a
fire-sausage (FS) 15) of height R = 2R 2U and width P'T 0 1), containing
stationary "quark waves" of momentum Ikl '" M/2 moving with small
divergences [0 '" RT/R2M] along the direction ~o.
Why do we choose these seemingly arbitrary superpositions of

states? The reason is just because the ~articular structure of the
FS is preserved in the process by which stable hadrons are produced
in the observed final states 1S ).
Let us in fact suppose that we produce in some way (which does

not interest us here) one of the high-mass meson states. Such a state
Rl (M)
~
..
~~ ~I~
~ no
2R2M
Fig. 10 The space structure of a fire-sausage of mass M

is not stable, but it has a finite probability amplitude for de-

caying into two other such states of lower mass (see Section 5).
Calculation l5 ) shows that the configuration of the final state which
is by far the most probable, is the one in which one bag has a very
low mass (~) and the other a mass (M) which is a large fraction of
the original one. Thus the most probable (canonical) decay con-
figuration of the initial mesonic state (1) is given by the cascade
chain depicted in Fig. 11.
To find out the structure of the final state we must square

the amplitude of Fig. 11 and sum over all possible states MI , M2 ,
and integrate over the full phase space. In carrying out this cal-
culation we make two important observations:
a) ~n the square of the probability amplitude interferences for

MI I M{, M2 I M~, ... , turn out to be negligible, so we can
visualize the cascade process as a classical one;
b) fixing the mass of ~I the phase of the transition amplitude

for the various (quasi-) degenerate states of different angu-
lar momenta are fixed in such a way that the state Ml can be
considered a FS as in (8.2) along the direction of the momen-
tum of ~l.
From these observations it follows that also M2 , M3 , are

fire-sausages along the directions of ~2' ~3' From the struc-
ture of the FS and the fact that their interactions are simply given
Fig. 11 The "canonical" decay pattern

of a high mass meson (1)
770 G. PREPARATA
by space-time overlaps it follows very easily that the low mass

hadrons ~1' ~2' are produced with very low momentum transverse
to the direction of the initial FS.
By use of the observations (a) and (b) we can dramatically

simplify the treatment of the cascade decay of the FS and, as has
been shown in Ref. 15, it all boils down to solving an integral
equation of a very simple type 19 ), the solution of which shows the
following interesting characteristics.
i) Scaling
The invariant one-particle inclusive distribution
shows strong interaction scaling, i.e.
1(. _ 2~
) - ~ J
(8.4)
and the function f(x,PT) is peaked for small values of PT' the mo-
mentum component transverse to the FS direction.
ii) Logarithmic multiplicities
As a consequence of (8.4) the particle multiplicities increase

logarithmically with the FS mass, i.e.
(8.5)
iii) Short-range correlations
Owing to the peculiar character of the cascade, particle cor-

relations can only arise when the final stable hadrons emerge from
low mass resonances. This bars the possibility of having long-
range correlations from single FS decay. In the ideal case where
~1' ~2' are all TIIS, the distribution would be of a Poisson
type.
iv) CZustering
By inserting the meson wave functions, which are known ex-

plicitly, we can compute the ratios of multiplicities for "direct"
production of lOW-lying meson states. Calculation gives
(8.6)
These surprising results are the consequence of a fairly tricky

interplay of quark spins and meson masses, and show that even though
direct production of TI'S is strongly favoured by the multiperipheral
character of the dynamics of FS decay, quark spins and wave-function
factors end up "favouring" direct production of vector mesons and
higher resonances. This finding seems to be confirmed by very re-
cent observations and constitutes a derivation of the cluster model,
which seems to be very adequate in describing the short-range part
of particle correlations at high energy20).
From the previous properties it is clear that FS decay is a

strong candidate to account for the most important aspects of ha-
dronic final states. Thus we can set up the correspondence:
FIRE-SAUSAGES ~ FINAL STATES
In this way the extraordinary complication of high-energy in-

elastic processes reduces to the problem of finding the mechanisms
by which, in the different reactions, one produces a smal1 21 ) num-
ber of FS's which then decay most of the time according to the
canonical pattern just described.
8.2 High-Energy Scattering and the "Pomeron"
After having described how the FS concept emerges from the

spectrum and the space-time structure of high-mass meson states,
we must ask how FS's get produced in high-energy collisions. In
fact, we can conceive of inelastic high-energy scattering as a two-
step process:
initial state + FS's + final state (FS's decay) .

772 G. PREPARATA
p (8) cc (1+C05 2 8)
Fi0 12 The single FS production ~n e+e- annihilation
In our approach the natural place to look for the production

of a single FS is e+e- annihilation at high energy. He can visua-
lize this process as in Fig. 12; from the annihilation of the lep-
ton, through the intermediate heavy photon, a FS ~s created at an
angle e with probability pee) cr (1 + cos 2 e) due to the quark
spin ~ nature. The FS then decays according to the canonical pat-
tern described in the previous section, thus generating a "jet"
along the direction e. According to recent results from SPEAR and
DESy22), we can conclude that this description is strongly supported
by the experimental observations.
Thus to produce the many particle states which are so familiar

~n all high-energy collisions we need not produce a large number of
FS's. In fact, as already remarked, at the very root of the geo-
metrodynamical approach there lies the idea that once we have taken
into account the rich structure of hadrons inherent in their ex-
tended nature, the residual hadron-hadron interactions are quite
weak and can be treated perturbatively. Thus, in general, even at
the very high energies available at present machines we can visua-
lize the collisions as giving rise to a small number of FS's which
then decay into the final detected hadrons. For definiteness, let
us now consider TITI scattering at high energy. According to our
It It
It It
(1) (2 )
Fig. 13 (1) One- and (2) two-FS contributions

to the TITI cross-section
perturbative approach we consider to succeSS1ve orders the contri-

butions to high-energy inelastic scattering as reported in Fig. 13,
where the "black boxes" have their explicit representation in terms
of quark lines 8 ,9).
At high energy, the process (1) can be shown to die off like
1/s2, owing to form-factor damping 9 ) and can be neglected. Process
(2), however, can be shown 9 ) to yield a cross-section which 1n-
creases like R~. One can, in fact, compute
(8.7)
The production of 4, 6, ... FS's gives logarithmic corrections to

(8.7) and unless we go to extremely high energies they can be
treated as perturbations. It thus follows that the main mechanism
for producing highly inelastic states at high energy involves the
production and subsequent decay of two FS's. This means that
asymptotically the following relation holds between the multiplicity
in hadron-hadron scattering and e+e- annihilation 15 )
(8.8)
This relation has been checked by Ferbel and Stix 23 ) and found con-
sistent with data; its correctness constitutes a good support for
the idea that in the geometrodynamical framework "strong inter-
actions" can be .treated as perturbations.
774 G. PREPARATA
mass cluster
Fi~. 14 The production of two FS's 1n pp collisions
For proton-proton scattering the situation is a bit more com-

plicated owing to the fact, mentioned above, that baryonic FS's can
be neglected at high energies. Diagram (2) of Fig. 12 must now be
replaced by the one in Fig. 14, which shows that the dominant final
states in baryon-baryon scattering comprise two leading clusters
and two FS's. It 1S very interesting to note that it is precisely
the non-existence of baryonic FS's that forces the baryon number to
emerge in the final states always in the form of a low-mass cluster
which, for kinematical reasons, takes a fair fraction (~ ~) of the
incident baryon momentum. On the other hand, inelastic production
originates from the scattering of the virtual mesons left behind
by the leading clusters.
In this fashion a baryon interacting at high energy must neces-

sarily give off about half of its momentum to the leading cluster.
This structure leaves the interaction region simply as a spectator
of the collision, which involves thoroughly only the virtual meson
carrying the remaining fraction of the initial baryon momentum. All
this strongly suggests that the "missing momentum" of deep inelastic
scattering 24 ) rather than going into gluons goes into something much
less elusive, i.e. the leading cluster.
8.3 Particle Production at Large PT
After having discussed the le.ading mechanism of high-energy

particle production, we can now investigate the production of par-
ticles with large PT'
We have seen that the dominant mechanism for particle produc-

tion at high energy consists in the production of two FS's and in
their subsequent "canonical" decay into r.1any particles. Among the
features of the "canonical" decay we have quite a strong cut-off in
transverse momentum, which renders this decay configuration ex-
tremely inefficient (much more than experimentally observed) to
produce particles at large PT' Ue must then ask whether there are
non-canonical decay configurations of the FS which yield more abun-
dant production of particles carrying a large PT'
The answer, as was shown in previous work, is indeed positive.

The dominant mechanism is depicted in Fig. 15, which yields par-
ticles at large PT through the "tilting" of one of the FS's of the
decay chain at some finite angle 8.
The process can be described as follows: one of the two FS's

produced at high energy starts decaying canonically until at one
link of the chain one FS changes the direction of its axis by a
finite angle 8 and initiates another canonical decay process along
Fig. 15 The "tilting" Of a FS along the decay chain

776 G. PREPARATA
2:n ~ _M 2
tt= _M2/2 (1-C059)

Fig. 16 The diagram determining the "tilting" probability (8.9)
the new direction, thus yielding particles at large PT. The tilting
probability for a FS of mass M has been computed for scalar quarks
(see Ref. 9) from the diagram in Fig. 16, a similar calculation can
be carried out for spinning quarks with the result
(8.9)
where
\+* (.'+Gos&y"
l1-rA&)!.
and A} is a "coupling constant" whose value can be, with some fur-
ther labour, calculated. It is interesting to note that (8.9) is
similar, but not identical, to the t10ller scattering cross-section
for two quarks which has been invoked by people working with QeD
and the parton mode1 25 ). It has the same mass dependence (M- 4 ),
but a slightly different angular dependence, thus leading to results
which are similar to the parton model calculations. If we want to
compute the one-particle inclusive cross-section we must juxtapose
three separate pieces, as indicated in Fig. 17.
The tilting probability and the one-particle inclusive decay

of a FS has just been discussed, so in order to complete the oicture
we must find out what is the probability of producing a high-mass
qq pair (FS) in the hadronic high-energy collision. According to
the previous discussion, apart from over-all normalization the
first diagram in Fig. 16 can be decomposed according to Fig. 18.
+~+
p
( tilting )
amplitude
Fi~. 17 The elements needed to calculate

the one-particle large PT cross-section
leading cluster
leading cluster
Fig. 18 The inclusive production of a qq pair
In deep inelastic scattering we encounter a similar diagram

(see Fig. 19), so that it is not difficult to see that the "quark
distribution functions" measured in deep inelastic scattering
naturally appear also in the description of the diagram in Fig. 17,
in agreement with the postulates of the parton model.
Fi~. 19 The diagram giving rise to the

structure functions in deep
inelastic scattering
778 G. PREPARATA
An important question which has recently been drawing a great

deal of attention is related to the quark transverse momenta. In
the parton model it has always been assumed that partons carry only
a small transverse momentum, of the same order as the one observed
in hadronic collisions, i.e. (PT) ~ 300 MeV. According to recent
analyses, it appears that such a value should be more than doubled,
thus creating some difficulty for the parton model. How do we fare
in this respect? It should be noticed that the quark degrees of
freedom in this approach do carry even on the low-lying hadronic
states a momentum of the order of 1 GeV 26) and that on average
% of it is transverse. How about the small PT observed in single-
particle distributions at high energy? If we recall the result of
Section 2 that the great majority of TI'S are generated by resonance
decay, that also can be naturally understood. Thus in our approach
we have (PT) quar k ~ 700 MeV.
Putting everything together we arrive (after some simple cal-

culation) at the following expression:
2 dcr ~ A.
u: d3p ",4
T).
+ +H ....qt~) t~.. " l'X,)1 q(t.se'Jt) .fL~) = ~4 F(7;,6) (8.10)

T
where
are the qq distribution functions for the hadron H:
~ (CD)~) = (1+-c.o,&)z [1 +.L (I"t-Go)&)Z.1 (8.11)

1-cos~ +
and e* is the angle of the particle momentum in the qq centre-of-
mass; and finally fez) is the one-particle distribution function
in the FS decay which, according to Section 2, coincides with the
one measured at SPEAR and DORIS.
QUAR K-GEOMETRODYNAM ICS 779
The most remarkable aspect of (8.10) is the famous p~4 "scaling

behaviour", which was suggested a long time ago 27 ) , and confirmed
by several calculations. Although it is not supported by the ex-
perimental data, it seems hardly avoidable from the theoretical
standpoint, and failure to observe evidence of it at higher energies
and transverse momenta would constitute a serious difficulty for
the approach advocated here. Thus there should be another mechanism
which, even though asymptotically irrelevant, can in fact in some
intermediate PT range (PT ::; 6 GeV) dominate over the "tilting" pro-
cess. Indeed, the existence of such a mechanism was ascertained in
previous work 9 ), where it was shown that the direct decay of a FS
into a low-mass cluster and another smaller FS at large angle (see
Fig. 20) lead in fact to a faster drop for the inclusive cross-
section.
Going through the same steps as before, we can write:
2E ~~ \
'ftO~
= A~ ~
n CJq
r)
dX,dy. a t ~ M."~,,
<Jc.,) t H ..c
-.
lx,) +
T
.te.wi..,., Q
rT
... 1+1,..qlx,) t+4."CJ"-~')1 ~'(<DU~"") .f'l'Z.) = (8.12)
= Q (1p.
'(S)
e) .
[
n
Fig. 20 The next leading mechanism for

high PT particle production
780 G. PREPARATA
where
and f"(z) is a fairly complicated function of the particle frac-

tional momentum in the qq centre of mass 28 ).
Thus, depending on the value of A2 (which can in principle be

computed), we may have (S.12) dominate over (8.10) in some limited
PT range, and give the observed p~8 behaviour for the one-particle
inclusive cross-section.
To conclude, our theory of large PT processes predicts that

asymptotically the one-particle cross-section is described by the
formula (S.lO), and that in an intermediate PT and energy ranRe the
next leading contribution (S.12) may dominate the cross-section be-
haviour. The cross-section dependence on different beams and tar-
gets is practically the same as in the parton fusion model of
Landshoff and Polkinghorne 29 ). All this is nicely compatible with
present experimental knowledge.
S.4 Structure of Pinal States
In this last section I shall briefly report on an analysis of

the structure of final states arising in laree PT collisions done
in collaboration with Rossi 30 ). From the physical picture assumed
1n this report, it follows that the high PT events originate in the
two-step process
+l1++lZ ~ L,+L2. + l"S; + FSa
L. Q (JT) + 1="~ l-~ ) (S.13)
where L 1 ,2 are the leading clusters and R(PT) denotes a low mass
(~ 1 GeV) resonance which gets emitted at large PT through one of
the two mechanisms analysed in the previous section. It is quite
obvious that an analytical description of such complicated con-
figurations is out of the question; thus we decided to follow a
strategy based on generating high PT events on the large CERN
computer using Monte Carlo methods. A detailed account of our

method is reported in Ref. 30, where we show that several of the
features observed in the CERN ISR experiments 31 ) are naturally re-
produced.
Let me conclude this section by stressing that this methodology

is likely to be the most efficient one to analyse and to compare
with theory the amazing kinematical complexi.ty of the multiparticle
states produced in high-energy collisions.
9. CONCLUSIONS
In these lectures I have tried to give a presentation, as ex-

tensive as possible, of the new approach to hadrons and their inter-
action which has been called Quark-Ceometrodynamics. I hope I have
convinced you that this theory is capable of answering a great num-
ber of questions related to hadronic structure and high-energy be-
haviour. The principles are clearly stated, but the theoretical
framework still suffers from some residual flexibility which could
be removed by means of "theoretical experimentation". By this I
mean that there are some aspects like three-bag overlaps and high-
energy couplings whose precise quantitative evaluation needs further
analysis in order to fix some of their structural elements. Only
after this is done will one be able to make definite predictions
on the enormously complex field of particle production and decays.
Work is in progress and I hope to be able to report on it at

some future Erice School.
782 G.PREPARATA
REFERENCES AND FOOTNOTES
1) The naIve quark model was introduced in 1965 by G. Morpurgo,

Physics 2, 915 (1965) and later by R.H. Dalitz, in Proc.
13th Internat. Conf. on High-Energy Physics, Berkeley, 1966
(Univ. of Calif. Press, Berkeley and Los Angeles, 1967),
p. 215.
2) R.P. Feynman, Photon-hadron interactions (W.A. Benjamin, New
York, 1972).
3) Colour was first introduced by O.W. Greenberg, Phys. Rev.

Letters 13, 598 (1964); and later developed into QCD by
H. Fritzsch, M. Gell-Mann and H. Leutwyler, P~ys. Letters
47B, 365 (1973).
4) See the discussion of K.G. Wilson, Proc. 1975 International

School of Subnuclear Physics, Erice, Italy, in New phenomena
in subnuclear physics, part A (ed. A. Zichichi) (Plenum
Press, New York, 1977), p. 127.
5) The original paper on the MIT-Bag is by A. Chodos, R.L. Jaffe,

K. Johnson, C.B. Thom and V.F. Weisskopf, Phys. Rev. D 9,
3471 (1974); its extension is due to J. Kuti and colla~
borators; see, for example, J. Kuti, Proc. of the European
Conference on Particle Physics, Budapest, 1977.
6) This approach has been proposed by G. Preparata and N. Craigie,

Nuclear Phys. B102, 478 (1976). Among its precursors the
works of Yukawa and his group in the early fifties are the
most noteworthy.
7) V.F. Weisskopf, Proc. 1975 International School of Subnuclear

Physics, Erice, Italy, in New phenomena in subnuclear phy-
sics, part A (ed. A. Zichichi) (Plenum Press, New York,
1977), p. 241.
8) G. Preparata, Proc. 1975 International School of Subnuclear

Physics, Erice, Italy, in New phenomena in subnuclear phy-
sics, part A (ed. A. Zichichi) (Plenum Press, New York,
1977), p. 193.
9) G. Preparata, Proc. 1976 International School of Subnuclear

Physics, Erice, Italy, in Understanding the fundamental
constituents of matter (ed. A. Zichichi).
10) J.A. Wheeler, Geometrodynamics (Academic Press, New York, 1962).
11) I have in mind the remarkable posthumous book by H. Hertz, The

principles of mechanics presented in a new form (Dover, New
York, 1956).
12) For a discussion of the U(l)-problem, see S. Coleman's lec-

tures at this School.
13) G. Preparata and K. Szego, Phys. Letters 68B, 239 (1977)
14) F. Csikor and G. Preparata, Geometrodynamics for quarks and

hadrons; the definition of currents, CERN TH 2396 (1977) .
15) N. Craigie and G. Preparata, Nuclear Phys. Bl02, 497 (1976) .
16) I am following closely the presentation in G. Preparata, Proc.

8th Internat. Symposium on Multiparticle Dynamics,
Kaysersberg, France, 1977 (C.R.N., Strasbourg, 1977), p. B-65.
17) This follows from the particular structure of baryonic wave

functions; see Section 4.
18) This is a consequence of our normalization procedure; see

Ref. 9.
19) This equation has been introduced first in A. Krzywicki and

B. Petersson, Phys. Rev. D~, 2606 (1972).
20) For a recent review of the situation, see A. Wrobleski, Proc.

8th Internat. Symposium on Multiparticle Dynamics,
Kaysersberg, France, 1977 (C.R.N., Strasbourg, 1977), p_ A-I.
21) The smallness of the number of produced FS's is a consequence

of the perturbative nature of hadronic interactions (see
Section 5).
22) For a recent review, see the excellent lectures of B. Wiik and
G. Wolf, Proc. Ecole d'ete de physique theorique 1976, Les
Houches, France, Weak and electromagnetic interactions at
high energy (North-Holland, Amsterdam, 1977), p. 403.
23) P. Stix and T. Ferbel, Phys. Rev. D 11, 358 (1977).
24) As is well known, this fact has been taken as evidence for the
existence of gluons.
25) See, for example, S.J. Brodsky, Proc. 8th Internat. Symposium
on Multiparticle Dynamics, Kaysersberg, France, 1977 (C.R.N.,
Strasbourg, 1977), p. B-18l.
26) See Sections 3 and 4.
27) S.M. Berman, J.D. Bjorken and J. Kogut, Phys. Rev. D~, 3388
(1971) .
784 G.PREPARATA
28) An approximate calculation of f'(z) is contained in Ref. 15.
29) See the review of J. Polkinghorne, Proc. of the European

Conference on Particle Physics, Budapest, 1977.
30) G. Preparata and G. Rossi, Nuclear Phys. BIll, III (1976).
31) For a review, see P. Darriulat, Proc. Internat. Conf. on High-

Energy Physics, Tbilisi, USSR, 1976 (Dubna, USSR, 1977),
p. A4-23.
DIS C U S S ION
CHAIRMAN: G. Preparata
Scientific Secretary: M. Bate
DISCUSSION No.1
COLEMAN:
You stated that your theory involves a minimization

condition. I ask about it because if you impose non-
linear conditions you are liable to run into difficulties
with the principle of superposition.
PREPARATA:
The solutions can be factorized, in the rest frame,

into a function of x times a function of t; space and
time are decoupled and so there is no problem at the
boundary. Minimization takes that solution of all the
time excitations which is closest to the free motion
solution. Higher time excitations, which lead to prob-
lems, are in this way avoided.
COLEMAN:
Is there only one admissible solution for every

momentum?
PREPARATA:
Yes.
785
786 DISCUSSION
COLEMAN:
So you are saying that all your solutions have the
same time dependence and the problem is effectively lin-
ear, and so there is no problem with superposition.
PREPARATA:
No problem.
HINCHLIFFE:
Is there any reason why your ml,m2 parameters are

the same as in the non-relativistic quark model?
PREPARATA:
I view it as a coincidence.
HINCHLIFFE:
What do you do with all the new states described by
Perez-y-Jorba?
PREPARATA:
In fact I have sufficiently many states in this
region. Concerning the 1500 PP state I believe the 3
years old explanation of the Russian group of Shapiro
et ale giving a very narrow width and correct mass
might be the right explanation.
CELMASTER:
Are we to think of your wave functions as Bethe-
Salpeter wave functions? What is their meaning?
PREPARATA:
I will explain more about this in my next lectures.
For now, the wave functions contain information about the
momentum and space-time distribution of quarks inside the
hadrons but the details depend on the probe used.
DISCUSSION 787
BIGI:
What is the difference between your model and the

MIT-bag, for example, as far as the spectrum is concerned?
PREPARATA:
I have much fewer states since there are no excita-

tions connected with the surface.
CHEN:
I presume the relation radius CCmass is worked out

mainly to explain the Chew-Frautschi plot. Does this
mean that, for example, the radius of the~is 3 times
that of the It ? What does it say about the shape of high
spin particles? What is the relation between the charge
radius and the strong interaction radius in your model?
PREPARATA:
The relation R ( M derives from the principle of max-

imum freedom. It is a quite welcome coincidence that it
gives such a nice Chew-Frautschi plot. As for the 'l.-7C
radius difference it turns out quite small; the mass-
radius relation is only an asymptotic one. For high spin
particles, which are also heavy the radius will be prop-
ortional to their mass. As for the relation between the
strong interaction and the charge radius it has to do
with both the wave functions and the scattering mechan-
isms, and cannot be given in very simple terms.
DISCUSSION No.2 (Scientific Secretary: E. Mottola)
REITHLER:
You quoted the values of the three mass parameters,

mp = 0.100 GeV, m" = 0.480 GeV and mc = 1.86 GeV as in-
puts to the model. Where are they taken from and what
are the uncertainties in these values?
788 DISCUSSION
PREPARATA:
The masses come from the fit of the mass formula,
M = El + E2 - W with Ei =i cr,
particle spectrum of 'P, 6:>,
k 2 + mi to the physical
etc. The resulting values
of the mass parameters have very small uncertainties.
MOTTOLA:
Please explain in more detail the meaning of the

diagrams you draw for the unitarity corrections and scatt-
ering amplitudes.
PREPARATA:
The diagrams have a meaning in close analogy with

ordinary Feynman graphs. Each line corresponds to a
quark propagator, provided we remember that this is an
unusual propagator because the quarks cannot escape the
space-time bag. The vertex
?~:
is a non-local function ~i(P,x,y) unlike the usual local
vertices in Lagrangian field theory. This function comes
from our solution for the quark orbits in the bag. For
example,
~'~
corresponds to ).16JdX dy dz q>1(X,Z,P1) 'P2 (x,y,p,,).
CP~(y,z,PS) or ~Sdk ce(p. ,k) CP2(p:l.,k+P.:a /2 ) 'f3(P3,k-p:a./2)
~ {PI -p. -p) in momentJm'space. What we have is just
the ove~la~ integral for non-local vertices analogous to
local ~~field theory. Diagrams such as
correspond to
DISCUSSION 789
'3
in cp-theory and:
is analogous to:
-0-
with the non-local vertex functions substituted. This
last diagram is a self-energy correction or unitarity
diagram which must be perturbatively taken into account
to give mass shifts (real part) to the zeroth order mass
values and decay rates (imaginary part). If the frame-
work of the theory is reasonable, then these corrections
will be small.
CELMASTER:
What about renormalization? Are any of your integ-

rals infinite?
PREPARATA:
In the geometrodynamical diagrams infinities only

occur when we sum over the particle spectrum (the index,
i, above) which, as I have shown in my lectures, is in-
finite. If we analyze in detail the nature of these
divergences, we see that they are equivalent to the ones
occurring in a theory with normal quarks and gluons. I
do not have time to elaborate on this point but I shall
only say that the power counting of this theory is the
same as, say, QED.
BIZZARRI:
You mentioned the possible existence of an A2~3~

decay as distinct from the A2~ t~ decay. It seems to
me that you shall always have a It ie-pair in a P-wave and
therefore, that this decay will not be distinguishable
from the p1L decay.
PREPARATA:
Yes, you are absolutely right. The X~ pair can

mimic the p meson decay. They will be indistinguishable
unless I calculate very carefully the mass spectrum
790 DISCUSSION
distribution of the pions. The result for the total A2

decay rate involves the sum of the contributions you
have mentioned.
CELMASTER:
Does your model satisfy the usual requirements of

unitarity, crossing, etc.?
PREPARATA:
The program of calculations that I have outlined

graphically shows how to implement unitarity. However,
although this procedure is well-defined, the corrections
must be small or else the picture is physically unsatis-
factory. Crossing is also an elementary property of the
diagrammatic formulation given. The question of analyti-
city is a much more difficult one. For scattering ampli-
tudes everything looks OK but there are some difficulties
in the form factor analyticity properties. However,
these are common to all confinement theories and may not
be fatal to this one. The analysis is far from complete
and we are continuing to investigate this question.
CELMASTER:
If you are right about the charmonium D and S states

m1x1ng, i.e. "'II 'mixing with the ~D1 state, in order to ex-
plain 'P widths, shouldn't we have already seen the 3D1
in e+e- annihilation experiments?
PREPARATA:
My view is that the 't' is a mixture of an Sand D

state and that the bump at 4.028 GeV is the other mixture.
There is a possible difficulty with the state at 3.775
GeV observed at SPEAR, which decays into Dfi. My belief
is that this is a "nuclear-like" state, i.e. it is a DD
state weakly bound by a IL or p exchange force but detail-
ed calculations have not yet been carried out.
DEO:
In your model how do you account for the existence

of 2 and 3 quark bound states but not states of larger
DISCUSSION 791
numbers of quarks, such as qqqq?
PREPARATA:
I have made the colorless or triality-zero ansatz

and invoked the principle of simplicity which excludes
such exotic states. If color states are found, the
model will have to be changed, if not completely aband-
oned. The baryonium states of 4 quarks may possibly be
included in the model but we should first be sure that
simpler explanations are experimentally ruled out.
DEO:
How does the interaction you have described follow

from geometry'2
PREPARATA:
The interaction is simply proportional to the geo-
metric overlap of the wave functions of the initial and
final states.
DEO:
Capri and Chiang have considered a model very S1m1-

lar to yours, and, in particular, they have also consid-
ered the question of unitarity mentioned before. How
does their model differ from yours?
PREPARATA:
I have not seen their recent work but I believe that
their approach is very similar to mine, as they began the
work after my visit to Munich where I gave a talk on this
model.
BIGI:
You said that in the baryon wave function, the re-

lative coordinates ~and yare parallel; yet this leads
to a configuration where the quark coordinates are dis-
tributed along a line with non-zero distances between
them. This is different from a configuration where two
792 DISCUSSION
quarks are together at one end and the third is on the

other end. So, don't you find more excitations in your
model than one finds in a model where the baryon is a
quark-diquark system?
PREPARATA:
No, because the important point is that one degree

of freedom, corresponding to the diquark angular momentum
is lost, i.e. it is zero. This is just as if we had a
quark-diquark system to start with - at least as far as
the number of excitations is concerned.
BALDINI:
You gave us your predictions on the r-recurrences.

What about the c.p-recurrences?
PREPARATA:
Recall the zeroth order approximate predictions of

the S and D states for the <f
.t ~ D e. S- D
n n
1 1.00 I. ~<6 1 D.76
cp: 2. 2.02 '2.62 ~: '2 1. 'b5
'3 2.E:.f!,
2.54
~
The states connected by the arrows can mix.
MIETTINEN:
What about the 1497 MeV narrow state and others ob-
served at Adone? You would not associate all of them
with <f' is that right?
PREPARATA:
My preferred explanation is that of the Russian

group which predicted a 1-- mainly D-wave state at 1515
MeV. The other states, if they exist, may possible also
be explained in a similar way.
DISCUSSION 793
LIPKIN:
I would like to point out that this ~'at 1500 MeV

looks strange from another point of view. It is really
degenerate with the f' and by SU(3), one should also
find a p'near the f at 1200 MeV but this p'has not been
found.
I also have a question on Dr. Preparata's model.
The standard quark model predicts the existence of baryon
excitations with both internal orbital angular momenta
excited, but also predicts that they will not be observed
in the single quark excitations seen in phase shift anal-
ysis. If such excitations are in fact found by product-
ion experiments, will this destroy your model?
PREPARATA:
If these states are found, then the baryon spectrum

cannot really be computed as I have set things up in this
model, where the one angular degree of freedom vanishes.
It would certainly be a serious difficulty and might even
be fatal to the model.
DISCUSSION No.3 (Scientific Secretary: W. Celmaster)
BERLAD:
Is it not trivial to get asymptotic freedom, since

the quarks are assumed to be quasi-free (principle of max-
imal freedom) at small distances?
PREPARATA:
The zero order theory is a free theory indeed. Only

when corrections are introduced, asymptotic freedom be-
comes non-trivial. A typical correction to deep-inelastic
scattering is
794 DISCUSSION
In the ultraviolet region each sUbdiagram of the type
~ -'> s..e..n So
("Pomeron" exchange) is equivalent, as far as power
counting is concerned, to "gluon-exchange" in a spinor
theory. So the theory looks renormalizable, but I can-
not yet make this a quantitative statement.
BERLAD:
A striking feature of large Pol. hadronic production

is the rise of the cross section at fixed p~ with energy.
How does your p~aterm behave with respect to energy?
PREPARATA:
This approach reproduces the parton model scaling,

i. e.
w'nen!. 1'2.
C1-x.L ")
Because .f(x",') is a rapidly decreasing function of X..1..)

then at fixed 1'.1.> d~CS"/d1>3 will be rising with energy.
CELMASTER:
Could you please explain your derivation of the

normalization factor for the fire-sausage wave function.
PREPARATA:
This analysis is carried out in detail in CERN TH-

2180. There is a normalization equation for the charge
form factor. At 0 momentum transfer
1 _
DISCUSSION 795
This is an approximation but all things are under control.

We use p-dominance (which is shown). The first diagram
is cal,led Wp' Then we have
1 ::. N'2.(M) ['+.I p -+ ~Y'lst J

You have to do the loop integral for 'W f and you have to
integrate over complex momenta because you are in the
unphysical region. So you do Gaussian integrals and ex-
tend your result to the unphysical momenta. Then you
find 'IJ~ /\111M e~p t..e.Mp/fo'\). This comes from the analytic
continuation of certain spherical harmonics such as
~J. (~ 1+ M~/M:z."). SO N 2 _ 1 / (1-t 11M ext> ltMf'/M) ")
and when ~ '> Y2 12.LM we see that there is a large supp-
ression. So 2 particles interact only when their impact
parameters are below a certain value, which is R~. If
this suppression didn't occur then by going to high en-
ough 1, i.e. on the leading trajectory, we could have
interaction for any impact parameter, and that could go
all the way to {S. That would be disastrous, for it
would be an interaction which wouldn't have a logarith-
mically increasing range. This all seems miraculous. It
made me very happy when I saw this come out. Notice, the
transverse dimension of the fire-sausage does involve a
typical strong interaction parameter which is the mass
of the p-meson. If we apply these considerations to the
J'/,-\, system we find that the cross-section of p compared to
'J"/,-\, goes like (m~/mf,), which is about 10 in accordance
with experimental data.
CELMASTER:
Does it not appear that this first order perturbation

from where y6u derive your fire-sausage shape is really
rather large to be treated as a perturbation?
PREPARATA:
Yes, it does appear that way, but it turns out that

the higher order corrections are really smaller.
JENNI:
In conventional parton models people have explained

the existence of single high transverse momentum photons
796 DISCUSSION
from high energy collisions by Bremsstrahlung in quark-

quark scattering. How are single photons described in
your firesausage picture and have you any measurable
predictions of this process?
PREPARATA:
To answer the last question first, I do not yet

have predictions. Photons are coupled to matter via
qq or vector mesons:
At high energy the direct coupling seems to dominate.

The photons can be produced at each link during the
firesausage decay.
DEO:
Can you explain the general principle for the form-

ation of the firesausage by summing up of spherical
harmonics up to a maximum l? Is there a general theorem?
If you cut off at .e~x even in virtual states do you
still maintain usual invariance properties of amplitudes?
PREPARATA:
You can assume all harmonics are in the same phase

and all ~>.eo) where (0= RJ.M/2, decouple. This gives
the firesausage. There are no problems in doing this.
DEO:
How does your geometry determine Yukawa coupling?

Do you introduce another coupling constant? You may
still have to introduce Acp4- coupling with another
undetermined strength.
PREPARATA:
I have introduced a minimum number of coupling con-

stants. I thought that Atp3 would suffice but then I
could not explain the pomeron so I had to introduce A~4
Then I stopped introducing new couplings. But this is no
DISCUSSION 797
different than in other field theories. No one tells us

the coupling constants of the gauge theories or the
masses of the Higgs mesons.
DEO:
Can you explain the physical significance of the

rainbow diagram?
PREPARATA:
It is the topology of linear chain decay. From the

space time structure you can show there are no crossings
between lines. You sum over the links and then take the
square. Because of the lack of interference you get the
integral equation. That enables you to find the spectrum
of particles coming out of linear chain decay.
HINCHLIFFE:
Why can't you have such a crossing of lines?
PREPARATA:
This is an interference term and will be greatly

suppressed due to the tremendous amount of coherence in
the firesausage. The decay is stepwise and the short
range correlation makes the interference small.
BIGI:
Don't these coherence effects you were just talking

about generate long range correlations in rapidity space?
PREPARATA:
No, exactly these coherence effects lead to cancell-

ations of interference terms and generate only short
range correlations.
BERLAD:
It may be true that cross diagrams are suppressed

but you can know that only after calculation. One must
798 DISCUSSION
count the number of interference diagrams.
PREPARATA:
Yes, but here the suppression is exponential.
WIGHTMAN:
Suppose Deo had asked "what if there are phases in

the sum defining your firesausage?" How would you just-
ify your derivation?
PREPARATA:
The phases were selected in such a way that the

firesausage structure was preserved in the decay chain.
If one changes them one gets neither a firesausage nor
a structure which gets preserved in the decay. But of
course in the end the result will be the same.
ATWOOD:
One of the strongest constraints on dynamical theo-

ries of nucleon structure will be its confrontation with
measured form factors and structure functions. Specif-
ically, have you made calculations of the elastic form
factors, 6e: and <4M or of the inelastic structure funct-
ions W1 and w~ for protons and neutrons, suitable for
compar~son to the existing experimental data?
PREPARATA:
No, not yet, because the baryon problem has just

been tackled recently.
ATWOOD:
By suitable, I mean for example does your model give

the 1/ ~4 behaviour ofGMfor the proton observed in the
data and predicted by Brodsky and Farrar using the Con-
stituent Interchange Model?
DISCUSSION 799
PREPARATA:
Yes, of course, the diagram ~ will give
1/Q+. My big problem is to see what happens to the vector

dominated piece. But this is a problem confronting all
other workers.
KRIPFGANZ:
This is a follow-up question. You have given an al-

ternative explanation of the right hand side of the mom-
entum sum rule in deep inelastic scattering without in-
troducing gluons. In this case, it is important to look
for other experimental possibilities to distinguish be-
tween your scheme and normal quark-gluon models. A sen-
sitive experimental quantity seems to be the ratio of
neutron to proton structure function at large x. Exper-
imentally this ratio is close to 0.25. QeD-based calcul-
ations (G. Farrar et al.) yield 9/21, as far as I remember.
Do you have a quick estimate of your prediction? A very
naive first guess seems to give a value of 1.
PREPARATA:
I have no way at this point of supporting any numb-

ers. For this, detailed calculations have to be done.
For instance, one might naively expect that because the
nucleon couples only to mesons, there should be as many
quarks as antiquarks. My answer to that is that you have
to sum over all the spectrum. There is a big coherence
effect in the region near x == 1.
MOTTOLA:
Why do you introduce the vector meson coupling to

photons in addition to the direct quark coupling? Is it
theoretically required by your principles or needed to
reproduce data on form factors, etc.? and why only one
meson - what about the many other possibilities?
PREPARATA:
Photons must have an amplitude to go into mesons by

general quantum arguments and from experimental data. So,
in that sense, it is both theoretically required and
800 DISCUSSION
phenomenological in or~g~n. The direct quark coupling

is not realistic experimentally nor is it theoretically
reasonable to assume the amplitude for N v~rtual-+meson
is o. I find that the other mesons are much less import-
ant according to standard dominance arguments - though,
of course, in principle they all contribute.
ATWOOD:
One comment: following Brodsky, if you plot G~/Q4

for the measured data you find it is flat from Q~ 5
GeV'Z. to the highest measured QI. of 33.4 GeV~.
PREPARATA:
Yes, and this is even more puzzling. No one under-

stands it.
WEILER:
Your baryon wave functions contain a quark and a
diquark. So if you impose Fermi statistics on your sol-
utions the u and d quark components of the wave function
will have different spacial properties, leading to
F:n / t:..ep (X.1
\ r 6'I
'\) ...J. fT
as observed in experiment. Have you
tried to explicitly include Fermi statistics?
PREPARATA:
I do include colour and Fermi statistics. As far

as the rest of your question goes this is an involved
problem and I cannot say anything at this time.
WETZEL:
In the two couplings to the photon is there no

double counting?
PREPARATA:
The analyticity properties of the two diagrams are

completely different so there is no double counting.
DISCUSSION 801
WETZEL:
Have you started to think about neutrino scattering,
and coupling ~p. to hadronic states?
PREPARATA:
This just involves another axial current and won't
change the basic structure. It will just add another
kinematical form factor.
MARTIN:
Where do you get an entire function?
PREPARATA:
You get an entire function in the form factor be-
cause you cannot cut the quark lines. However for real
9~ the form factor could decrease very fast.
MARTIN:
However, unitarity corrections will produce a cut.
PREPARATA:
Yes, but this is a higher order effect. The point
like coupling is a bad thing. In deuteron it is well
understood because you have the two-nucleon threshold.
You can think of the situation here as if the q mass
goes to infinity. So you have a non-trivial 'i 2 behaviour
but no singularity. This is, I think, a very crucial
problem which everyone should be confronted with.
RAJPOOT:
Can one do weak interactions with your model? If
the answer is yes then does the idea of a firesausage
somehow turn into something equivalent to the inter-
mediate bosons?
802 DISCUSSION
PREPARATA:
Yes, it is possible to do weak interactions provided
one invents weak bags on similar lines to the theory
stated in my lectures. To implement this idea one would
have to alter the scale so that Ro goes down by a factor
of about a thousand.
WETZEL:
I come back to the SU(6) baryon multiplets: If I
understood you correctly the reason that the 20 dimen-
sional multiplets are absent lies in the condition that
the two quark coordinates have to be parallel. Now this
is the condition which has to be imposed for the solution
of the free approximation of the wave function. What
happens to the actual wave function?
PREPARATA:
The same condition also holds for the actual wave
function because it comes from the spatial part of the
differential equation.
WETZEL:
My second ~uestion is about the calculation of the
decay constant f~ , which is in good agreement with ex-
periment. The definition of f~ involves the pion mass
which presumably you cannot reproduce easily. Does
this affect f. ?
~
PREPARATA:
No. The value of m
TC.
drops out from the kinematics.
WETZEL:
Could you write down the qq wave function which
represents the state of the firesausage produced in
e+e- annihilation?
DISCUSSION 803
PREPARATA:
T.he wave function involves what I call a fat delta
function &~ j something which becomes a delta function
as R goes to infinity and it has a width of l/R. Then
~ = ZLo l2e+ l)
..eeO
~ (cps~) " ~ R:'....
z(pZ_m~") bo 2 C. ?,,~m~ )
1
e is the direction of the quark momenta with respect to
the firesausage axis.
WETZEL:
Should the angular momentum sum not be limited to
1= 1?
PREPARATA:
Yes, of course, but what this sum tells us is that
we must proj ect out the 1= 1. state. You produce your
firesausage as soon as your vector meson has emitted a
pion. The first step of the chain is determined by the
l=i projection probability but after that there is no
restriction on the angular momentum of the firesausage.
THE USES OF INSTANTONS
Sidney Coleman
Lyman Laboratory of Physics, Harvard University
Cambridge, Hassachusetts 02138
I. INTRODUCTION
In the last two years there have been astonishing developments
in quantum field theory. He have obtained control over problems
previously believed to be of insuperable difficulty and we have ob-
tained deep and surprising (at least to me) insights into the
structure of the leading candidate for the field theory of the
strong interactions, quantum chromodynamics. These goodies have
come from a family of computational methods that are the subject
of these lectures.
These methods are all based on semiclassical approximations,
and, before I can go further, I must tell you what this means in
the context of quantum field theory.
To be definite, let us consider the theory of a single scalar
field in four-dimensional Minkowski space, with dynamics defined by
the Lagrangian density
For classical physics, g is an irrelevant parameter. The easiest

way to see this is to define
' = g (1. 2)
805
806 S. COLEMAN
In terms of <p',
tJ.. = g\ (!- a]..l <P' a]..l <P' - ~ m2 <P' 2 - <P' If) (1. 3)
Thus, g does not appear in the field equations; if one can solve
the theory for any positive g, one can solve it for any other posi-
tive g; g is irrelevant. Another way of seeing the same thing is
to observe that, in classical physics, g is a dimensionful param-
eter and can always be scaled to one.
Of course, g is relevant in quantum physics. The reason is
that quantum physics contains a new constant, 11., and the important
object (for example, in Feynman's path-integral formula) is
;,.., (~a <P' a]..l<p' + ... ) . (1.4)

g-ll ]..I
As we see from this expression, the relevant (dimensionless) param-

eter is g2~, and thus, semiclassical approximations, small-~ ap-
proximations, are tantamount to weak-coupling approximations,
small-g approximations.
At this point you must be puzzled by the trumpets and banners
of my opening paragraph. Do we not have a perfectly adequate small-
coupling approximation in perturbation theory? No, we do not;
there is a host of interesting phenomena which occur for small
coupling constant and for which perturbation theory is inadequate.
The easiest way to see this is to descend from field theory
to particle mechanics. Consider the theory of a particle of unit
mass moving in a one-dimensional potential,
L = ~ ~2 - V (x; g) , (1.5)
where
1
V(x;g) = 2"" F(g x) , (1.6)
g
and F is some function whose Taylor expansion begins with terms of

order x 2 . Everything I have said about the field theory defined
by Eq. (1.1) goes through for this theory. However, let us con-
sider the phenomenon of transmission through a potential barrier
(Fig. 1). Every child knows that the amplitude for transmission
THE USES OF INSTANTONS 807
Figure 1
obeys the WKB formula,

1 (2
- - 1x dx'; 2(V-E)
IT (E) I = e -t1 I [1 + 0 (-ti)] , (1.7)
where Xl and x 2 are the classical turning points at energy E. This

is a semiclassical approximation. Nevertheless, transmission,
barrier penetration, is not seen in any order of perturbation
theory, because Eq. (1.7) vanishes more rapidly than any power of
-n, and therefore of g.
I can now make my first paragraph more explicit. There are
phenomena in quantum field theory, and in particular in quantum
chromodynamics, analogous to barrier penetration in quantum particle
mechanics. In the last two years a method has been developed for
handling these phenomena. This method is the subject of these
lectures.
The organization of these lectures is as follows: In Section 2
I describe the new method in the context of particle mechanics,
where we already know the answer by an old method (the HKB approxi-
mation). Here the instantons which playa central role in the new
method and which have given these lectures their title first appear.
In Section 3 I derive some interesting properties of gauge field
theories. In Section 4 I discuss a two-dimensional model in which
instantons lead to something like quark confinement and explain why
a similar mechanism has (unfortunately) no chance of working in four
dimensions. In Section 5 I explain 't Hooft's resolution of the
U(l) problem. In Section 6 I apply instanton methods to vacuum
808 S. COLEMAN
decay. Only this last section reports on my own research; all the
rest is the work of other hands. 1
I thank C. Callan, R. Dashen, D. Gross, R. Jackiw, M. Peskin,

C. Rebbi, G. 't Hooft, and E. Witten for patiently explaining large
portions of this subject to me. Although I have never met A. H.
Polyakov, his influence pervades these lectures, as it does the
whole subject. 2
A Note on Notation: In these lectures we will work in both
Minkowski space and in four-dimensional Euclidean space. A point
in Minkowski space is labeled x~, where ~ = 0, 1, 2, 3, and x O is
the time coordinate. In Minkowski space I will distinguish between
v
covariant and contravariant vectors, x~ = g~v x , where the metric
tensor has signature (+---). Euclidean space is obtained from
Minkowski space by formal analytic continuation in the time co-
ordinate, x"
-1X
0
A point in Euclidean space is labeled x ~ ,
where ~ = 1, 2, 3, 4. The signature of the metric tensor is (++++).
Thus covariant and contravariant vectors are component-by-component
identical, and I will not bother to distinguish between them.
Note that xy in Minkowski space continues to -x.y in Euclidean
space. The Euclidean action is defined as -i times the continuation
of the Hinkowskian action. When discussing particle problems, I
will use t for both Euclidean and Minkowskian time; which is meant
will always be clear from the context. In Section 2 explicit fac-
tors of -1l are retained; elsewhere, -tr is set equal to one.
II. INSTANTONS AND BOUNCES IN PARTICLE MECHANICS

2.1 Euclidean Functional Integrals
In this section we will deal exclusively with the theory of a
spinless particle of unit mass moving in a potential in one dimen-
sion: 2
H = ~ + Vex) (2.1)
We will rederive some familiar properties of this much-studied sys-

tem by unfamiliar methods. For the problem at hand, these methods
are much more awkward than the standard methods of one-dimensional
quantum mechanics; however, they have the great advantage of being

immediately generalizable to quantum field theory.
Our fundamental tool will be the Euclidean (imaginary time)
version of Feynman's3 sum over histories:
(Z.Z)
Both sides of this equation require explanation:

On the left-hand side, Ix i > and Ix f > are position eigenstates,
H is the Hamiltonian, and T is a positive number. The left-hand
side of Eq. (Z.Z) is of interest because, if we expand in a com-
plete set of energy eigenstates,
E In> , (Z.3)
n
then
<x Ie -HTft11 x. > = \' -E T/h <x In><n I x. > .
Len (Z.4)
f ~ n f ~
Thus, the leading term in this expression for large T tells us the
energy and wave-function of the lowest-lying energy eigenstate.
On the right-hand side, N is a normalization factor, S is the
Euclidean action 4
s = IT/Z dt ll(dX) 2 + vl , (Z.5)
-T/Z LZ dt J
and [dx] denotes integration over all functions x(t), obeying the
boundary conditions, x(-T/Z) Xi and x(T/Z) = xf . To be more
specific, if x is any function obeying the boundary condition, then
a general function obeying the boundary conditions can be written
as
x(t) = ~(t) + I cn x (t)
n
, (Z.6)
n
where the x 's are a complete set of real orthonormal functions

n
vanishing at the boundaries,
I T/Z
dtx(t)x(t)
-T/Z n m
<5
nm
(Z 7a)
x (T/Z) O. (Z.7b)
n
810 S.COLEMAN
Then, the measure [dx] is defined by

-k
[dx] = IT (2TI'tr) 2 dC n (2.8)
n
(This measure differs in normalization from the measure defined
by Feynman;3 this is why we need the normalization constant N.
However, as we shall see, we shall never need an explicit formula
for N.)
The right-hand side of Eq. (2.2) is of interest because it can
readily be evaluated in the semiclassical (small-h) limit. In this
case the functional integral is dominated by the stationary points
of S. For simplicity, let us assume for the moment that there is
only one such stationary point, which we denote by x,
oS (2.9)
ox
-= -
where the prime denotes differentiation with respect to x. Further,

let us choose the ~'s to be eigenfunctions of the second varia-
tional derivative of S at x,
An x
n
(2.10)
Then, in the smallJh limit, the integral becomes a product of Gaus-

sians, and we find
Ne -S (x) frr IT A~!:2 [1 + 0 (11.) ]

n
Ne- S (X)/lr [det(-a 2 + V"(X) ]-!:2[1 + OEtr)]

t (2.11)
(Of course, we are tacitly assuming here that all the eigenvalues
are positive. VIe shall shortly see what to do when this is not the
case.) If there are several stationary points, in general one has
to sum over all of them.
Equation (2.9) is the equation of motion for a particle of
unit mass moving in a Dotential minus V. Thus,
E 12 (ddXt) 2 - V (x) (2.12 )

v -v
------~~~-------+ X
( a) (b)
Figure 2
is a constant of the motion. This can be used to determine the

qualitative features of the solutions of Eq. (2.9) by inspection.
As a simple example, consider the potential shown in Fig. 2a.
Let us choose xi = x f = O. Figure 2b shows the inverted potential,
-V. It is obvious from the figure that the only solution of Eq.
(2.9) which obeys the boundary conditions is
x= 0 (2.13)
For this solution, S = O. Thus, from Eq. (2.11),
<Ole-HTfhIO> = N[det(-d~ + w2)]-~[1 + O(~)] , (2.14)

where
w2 = V"(O) (2.15)
In Appendix A, I show that, for large T,
(2.16)
Thus, the ground-state energy is given by
(2.17)
Also, the probability of the particle being at the origin when it

is in its ground state is
I<x = Oln = 0>1 2 = (w/7rn)~[l + O(~)] (2.18)
These are, of course, the correct semiclassical results. In

the small-~ limit, the particle is in a harmonic-oscillator ground-
812 S. COLEMAN
-v
v -0 o
x
-0 o
(a) (b)
Figure 3
state concentrated at the origin and its energy is the ground-state

energy of a harmonic oscillator.
2.2 The Double Well and Instantons
He now turn to a less trivial problem,5 the double well of

Fig. 3a. I will assume the potential is even, Vex) = V(-x) , and
will denote its minima by fa. As before, I will add a constant to
V, if necessary, to make V vanish at its minima, and I will denote
V"(a) by w2
We will attempt to compute both
(2.l9a)
and
<a \ e -HT\ -a> = <-a \ e -HT\ a>, (2.l9b)
by approximating the functional integral by its semiclassical limit,

Eq. (2.11). Just as before, the first step is to find solutions of
the classical Euclidean equation of motion, (2.9), consistent with
our boundary conditions.
Of course, two such solutions are those in which the particle
stays fixed on top of one or the other of the two hills in Fig. 3b.
However, there is another potentially interesting solution, one
where the particle begins at the top of one hill (say the left one)
at time -T/2, and moves to the top of the right hill at time T/2.
Figure 4
Since we plan eventually to take T to infinity, we will focus on

the form of the solution in this limit, where the particle attains
the tops of the hills at times plus and minus infinity. In this
case, we are dealing with a solution of the equation of motion with
vanishing E; whence
dx/dt = /2V (2.20)
X
Equivalently,
t tI + f dx'(2V)-~ (2.21)
o
where tI is an integration constant, the time at which x vanishes.
This solution is sketched in Fig. 4; it is called "an instan-
ton with center at tI". The name "instanton" was invented by
't Hooft. The idea is that these objects are very similar in their
mathematical structure to what are called solitons or lumps,6
particle-like solutions of classical field theories: thus the "-on".
However, unlike lumps, they are structures in time (albeit Euclidean
time): thus the "instant-". For the same reason, Po1yakov suggested
the name "pseudopartic1e", also used in the literature.
Of course, we can also construct solutions that go from a to
-a, simply by replacing t by -t in Eq. (2.21); these are called
"anti-instantons".
Two properties of these solutions will be important to us:
814 S. COLEMAN
(1) From Eq. (2.20), it is easy to derive a simple expression

for So' the action of an instanton (or anti-instanton)
a
So = fdt[~(dX/dt)2 + V] = f dt (dX/dt)2
dx/2V
(2.22)
= f
-a
Note that this is the same as the integral that appears in the
barrier-penetration formula, Eq. (1.7). We shall see shortly that
this is no coincidence.
(2) For large t, x approaches a, and Eq. (2.20) can be approx-
imated by
dx/dt = w(a - x) (2.23)
Thus, for large t, -wt

(a - x) ex e (2.24)
Thus, instantons are, roughly speaking, well-localized objects,

having a size on the order of l/w.
This is of critical importance, because it means that, for
large T, the instanton and the anti-instanton are not the only ap-
proximate solutions of the equation of motion; there are also ap-
proximate solutions consisting of strings of widely separated in-
stantons and anti-instantons. (You may be troubled by the sudden
appearance in the argument of approximate solutions, approximate
stationary points of S. If so, bear with me; I'll give a fuller
explanation of this point later.)
I shall evaluate the functional integral by summing over all
such configurations, with n objects (instantons or anti-instantons)
centered at tl ... tn' where
T/2 > tl > t ... > t > -T/2 . (2.25)

2 n
Figure 5 (next page) shows one such configuration. T is as-

sumed to be huge on the scale of the size of an instanton; thus the
smooth curves of Fig. 4 appear as sharp jumps on the scale of Fig.
5. (The vertical marks on the time axis will be explained shortly.)
Now for the evaluation:
(1) For n widely separated objects, S is nS o . This takes care
of the exponential of the action.
------~~-+--~L-~~------~t
T/2
Figure 5
(2) The evaluation of the determinant is a bit trickier. Let

.
us consider the time evo1utlon operator, e
-HT ,as a product of
operators associated with evolution between the points indicated by

the vertical marks on the time axis in Fig. 5. If it were not for
the small intervals containing the instantons and anti-instantons,
V" would equal w2 over the entire time axis, and thus we would ob-
tain the same result we obtained for a single-well potential in
Section 2.1,
-wT/2
e . (2.26)
The small intervals containing the instantons and anti-instantons

correct this formula. Thus we obtain
(.:t e --uJT/2 Kn , (2.27)
where K is defined by demanding that this formula give the right

answer for one instanton. Later we shall obtain a more explicit
expression for K.
(3) We must integrate over the locations of the centers:
T/2 t t l
f dt1f dt 2 f n- dt n = Tn/n! (2.28)
-T/2 -T/2 -T/2
(4) We are not free to distribute instantons and anti-

instantons arbitrarily. For example, if we start out at -a, the
816 s. COLEMAN
first object we encounter must be an instanton, the next one must be
an anti-instanton, etc. Furthermore, if we are to end up back at
-a, n must be even. Likewise, if we wish to end up at a, n must be
odd.
Thus,
<-a 1e -HTfh 1-a> [1+0(tr)] ,

(2.29)
while <ale-HT!ttI_a> is given by the same expression, summed over

odd n's. These sums are trivial:
<a 1e -HT/i11 -a> =

(w)~
mt e -WT/2 Hexp(Ke-S 0 f'r(T) t-exp(-Ke -S 0 fri.T)]
(2.30)
(From now on, to keep the page from getting cluttered, I will drop
the factors of [1 +0 (-h) ]; remember that they're omnipresent though
unwritten. )
Comparing this to Eq. (2.4), we see that we have two low-lying
energy eigenstates, with energies
E = Hrw -hKe -So /ff (2.31)
If we call these eigenstates 1+> and 1->, we also see that
1<+Ia>12 = 1<-Ia>12 = <al-><-I-a> = -<al+><+I-a> = !-(~)~

(2.32)
Of course, these are the expected results: the energy eigenstates
are the spatially even and odd combinations of harmonic oscillator
states centered at the bottoms of the two wells; the degeneracy of
the two energy eigenvalues is broken only by barrier penetration
(and thus the difference of the energies is proportional to the
barrier-penetration factor, e- Sofh), and the state of lower energy,
which we have denoted by 1->, is the spatially even combination.
Our next task is to evaluate K. Before we do this, though,
some comments should be made about what we have done so far:
(1) Really we have no right to retain the second term in Eq.
(2.31). It is not only exponentially small compared to the first
term, it is exponentially small compared to the uncomputed 0(1f2)
corrections to the first term. However, it is the leading contri-

bution to the difference of the energies, E - E ; a purist would
+
retain it only in the expression for this difference and not in the
expressions for the individual energies.
(2) Our approximation has been based on the assumption that
the instantons and anti-instantons are all widely separated. As a
consistency check, we should verify that the major portion of our
final result comes from configurations where this is indeed the case.
This check is easy to carry out. For any fixed x, the terms in
the exponential series, I xn/n! , grow with n until n is on the order
of x; after this point, they begin to decrease rapidly. Applying
this to the sum in Eq. (2.29), we see the important terms are those
for which
(2.33)
That is to say, for small 11, the important terms in the sum are
those for which niT, the density of instantons and anti-instantons,
is exponentially small, and thus the average separation is enormous.
Note that this average separation is independent of T; our approxi-
mation is indeed a small-n approximation; the conditions for its
validity are independent of T, as long as T is sufficiently large.
This approximation of summing over widely-separated in-
stantons is called the dilute-gas approximation, because of its
similarity to the approximation of that name in statistical
mechanics.
(3) Finally, I want to deliver the promised fuller explanation
of the idea of an approximate stationary point of S. Let us begin
by studying an integral over a single variable,
I f dt e-S(t)f!r
T
(2.34 )
o
where S is a function of t monotonically decreasing to some asymp-
totic value, S(oo). Thus the integrand has no stationary points in
the region of integration. Nevertheless, it is easy to find the
approximate form of the integral for small~ and large T:
818 s. COLEMAN
(2.35)
Speaking loosely, the integral is dominated by the stationary point

at infinity. It's straightforward to generalize this phenomenon to
multi-dimensional integrals: We assume an integrand whose graph has
a sort of trough in it; the line along the bottom of the trough
flattens out only as we go to infinity. Speaking less pictorally,
there is a line in the multi-dimensional space such that the inte-
grand is a minimum with respect to variations perpendicular to the
line and approaches some limiting value as one goes to infinity
along the line. Of course, the line could itself be generalized
to a hyperplane, a generalized ''bottom of the trough". This is in
fact the situation for our "approximate stationary points"; the
locations of the instantons and anti-instantons are the variables
along the bottom of the trough; S becomes stationary (and equal to
nS o ) only when they all go to infinity.
This concludes the comments; we now turn to the evaluation
of K.
We must study the eigenvalue equation, Eq. (2.10), with x a
single instanton. Because of time translation invariance, this
equation necessarily possesses an eigenfunction of eigenvalue zero,
(2.36)
[The normalization factor comes from. Eq. (2.22).] \.]ere we to inte-

grate over the corresponding expansion coefficient, c l ' in Eq. (2.6),
we would obtain a disastrous infinity. Fortunately, we have already
done this integration, in the guise of integrating over the location
of the center of the instanton in Eq. (2.28). The change of x(t)
induced by a small change in the location of the center, t l , is
dx = (dx/dt)dt i (2.37)
The change induced by a small change in the expansion coefficient,

cl ' is (2.38)
Hence,
(2.39)
Thus, in evaluating the determinant, we should not include the

zero eigenvalue, but we should include in K a factor 7 of (So/2~)~.
Hence, the one-instanton contribution to the transition matrix
element is given by
<aie-HTi-a>
one inst.
(2.40)
where det' indicates that the zero eigenvalue is to be omitted when
computing the determinant. Comparing this to the one-instanton
term in Eq. (2.29), we find
det(-a~ + w2 ) ~
K (2.41)
det'(-a t2 +V"(X)
This completes the computation.

Some remarks:
(1) To really sew things up, I should show that the formula
we have obtained for the energy splitting is the same as that ob-
tained by the traditional methods of wave mechanics. I do this in
Appendix B.
(2) I have been tacitly assuming that all the eigenvalues in
Eq. (2.10) are positive, other than the zero eigenvalue associated
with xl. It is easy to prove that this is indeed the case: It is
well-known that the eigenfunction of a one-dimensional Schrodinger
equation [like Eq. (2.10)] of lowest eigenvalue has no nodes, the
next-lowest eigenfunction has one node, etc. Because the instanton
is a monotone increasing function of t, xl' proportional to the time
derivative of the instanton, has no nodes. Thus zero is the lowest
eigenvalue and all the other eigenvalues are positive.
-~
(3) K is proportional to~ This factor came from the zero
eigenvalue associated with time-translation invariance. Later in
these lectures we will be analyzing theories that have larger in-
variance groups and for which the instantons have more than one zero
820 S. COLEMAN
eigenvalue associated with them. Clearly, for every zero eigen-

value ~here will be a factor of ~-~. This rule for counting powers
of ~ will be very important to us, for, as I explained in Section 1,
counting powers of ~ is equivalent to counting powers of coupling
constants.
2.3 Periodic Potentials
Let us consider a periodic potential, like the one sketched in
Fig. 6a. (For simplicity, I have chosen the minima of V to be the
integers.) If we ignore barrier penetration, the energy eigen-
states are an infinitely degenerate set of states, each concentrated
at the bottom of one of the wells. Barrier penetration changes this
single eigenvalue into a continuous band of eigenvalues; the true
energy eigenstates are the eigenstates of unit translations, the
Bloch waves. Let's see how this old result can be obtained by in-
stanton methods.
~
-1 o 2
(b)
Figure 6
As we see from Fig. 6b, the instantons are much the same as in
the preceding problem. The only novelty is that the instantons can
begin at any initial position, x = j, and go to the next one,
x j + 1. Likewise, the anti-instantons can go from x = j to
x j - 1. Otherwise, everything is as be fore.
Thus, when doing the dilute-gas sum, we can sprinkle instantons
and anti-instantons freely about the real axis; there is no con-
straint that instantons and anti-instantons must alternate. Of
course, as we go along the line, each instanton or anti-instanton
must begin where its predecessor ended. Furthermore, the total
number of instantons minus the total number of anti-instantons must
equal the change in x between the initial and final position eigen-
states.
Thus we obtain
~ (Ke -s 0 /-ti T ) n+il 8 _..

n!n! n-n-J++J_
(2.42)
where n is the number of instantons and n the number of anti-

instantons. If we use the identity
2'IT
8 ab = J d8 e i8 (a-b) /2'IT , (2.43)
o
the sum becomes two independent exponential series, and we find
1 2'IT
[ W )
'ITt!'
"2
e
-wT / 2
J e i (j - -'J+)8 21T
d8 -s "tf
exp[2KT cos 8 e ot]
o (2.44)
Thus we find a continuum of energy eigenstates labeled by the
angle 8. The energy eigenvalues are given by
E(8) = ~ hw + 2 hK cos 8 e -soAr (2.45)

Also,
(2.46)
Hearteningly, this is just the right answer.

822 S. COLEMAN
v -v
Figure 7
2.4 Unstable States and Bounces 8

Galilean pastiche:
SAGREDO: Let me test my understanding of these instanton
methods by studying the potential of Fig. 7a. I f I neglect barrier
penetration, in the semiclassical limit, this potential has an
energy eigenstate sitting in the bottom of the well. I wish to
compute the corrections to the energy of this state due to barrier
penetration. If I turn the potential upside down (Fig. 7b), Iob-
serve that the classical equation of motion has a solution in which
the particle begins at the top of the hill at x 0, bounces off
the classical turning point 0, and returns to the top of the hill
(Fig. 8). I will call this motion "the bounce". I will compute
the transition matrix element between x = 0 and x o by summing
over configurations consisting of widely separated bounces, just as
one sums over instantons and anti-instantons in the study of the
double well. Indeed, the sum is the same as that for the double
well (with the obvious redefinitions of So' w2 , etc.), save that
there is no restriction to an even or odd number of bounces. Thus
~ t
Figure 8
I obtain the complete exponential series, rather than just the odd
or even terms, and I find that
<0 Ie -HT/-t11 0> e

-WT/2
exp[KTe
-S I~
0 "] , (2.47)
and the energy eigenvalue is given by
(2.48)
SALVIATO: Alas, Sagredo, I fear you have erred in three ways.

Firstly, the term you have computed is small compared to terms of
order~2 which you have neglected, and thus you have no right to
retain it. Secondly, I see by your sketch that the bounce has a
maximum; therefore the eigenfunction xl' which is proportional to
the time derivative of the bounce, has a node. Thus it is not the
eigenfunction of lowest eigenvalue, and there must be a nodeless
eigenfunction, x o ' of a lower eigenvalue, that is to say, there
must be a negative eigenvalue. Thus K, which is inversely propor-
tional to the product of the square roots of the eigenvalues, is
imaginary. Thirdly, the eigenvalue you attempt to compute is no-
where to be found in the spectrum of the Hamiltonian, because the
state you are studying is rendered unstable by barrier penetration.
SAGREDO: Everything you say is correct, but I believe your
criticisms show how to save the computation. An unstable state is
one whose energy has an imaginary part; thus it is only to be ex-
pected that K should be imaginary. Furthermore, the term I have
computed, though indeed small compared to neglected contributions
to the real part of Eo, is the leading contribution to the imaginary
part of Eo. Thus the correct version of Eq. (2.48) is
(2.49)
where r is, as usual, the width of the unstable state.

As you can see, the Tuscan twosome are as quick-witted as ever,
although (also as ever) their arguments are sometimes a bit sloppy.
Sagredo has missed a factor of !; the correct answer is
824 S. COLEMAN
-------'=--t--------'~ t
Figure 9
(2.50)
To show that this is the case requires a more careful argument than
Sagredo's. The essential point is Salviato's observation that the
energy of an unstable state is not an eigenvalue of H; in fact, it's
an object that can only be defined by a process of analytic continu-
ation. I will now perform such a continuation.
To keep things as simple as possible, let us consider not an
integral over all function space, but an integral over some path in
function space parameterized by a real variable, z,
(2.51)
where S(z) is the action along the path. In particular, let us

choose the path sketched in Fig. 9. This path includes two impor-
tant functions that occur in the real problem: x(t) = 0, at z = 0,
and the bounce, at z = 1. Furthermore, the path is such that the
tangent vector to the path at z =1 is xo. Thus the path goes
through the bounce in the "most dangerous direction", that direction
with which the negative eigenvalue is associated, and z = 1 is a
maximum of S, as shown in Fig. 10. S goes to minus infinity as z
goes to infinity because the functions spend more and more time in
the region beyond the turning point, where V is negative; note that
this implies that Eq. (2.51) is hopelessly divergent.
Figure 10
If x = 0 were the absolute minimum of V, that is to say, if V

were as shown in Fig. lla, we would have, for the same path, the
situation shown in Fig. lIb, and there would be no divergence in
Eq. (2.48). Now let us suppose we analytically change V in some
way such that we go from this situation back to the one of interest.
To keep the integral convergent, we must distort the right-hand
portion of the contour of integration into the complex plane. How
we distort it depends on the details of the analytic passage from
one potential to the other. In Fig. 12, I have assumed that it is
distorted into the upper half plane. Following the standard pro-
cedure of the method of steepest descents, I have led the contour
along the real axis to z = 1, the saddle point, and then out along
a line of constant imaginary part of s. The integral thus acquires
v 5
___ _ _ _ _-+Z
x
(0) (b)
Figure 11
826 s. COLEMAN
Figure 12
an imaginary part; in the steepest-descent approximation,

l+ioo
imJ Imf dz(2TIh)-~ e-S(l}h
1
(2.52)
Note the factor of ~; this arises because the integration is over

only half of the Gaussian peak.
(If we had passed from one potential to the other in the con-
jugate manner, the contour would have been distorted into the lower
half plane, and we would have obtained the opposite sign for the
imaginary part. This is just a reflection of the well-known fact
that what sign you get for the imaginary part of the energy of an
unstable state depends on how you do your analytic continuation.)
Now, we have studied a one-dimensional integral, but we can
always reduce our functional integral to a one-dimensional in-
tegral simply by integrating (in the Gaussian approximation) over
all the variables orthogonal to our path. These directions involve
only positive or zero eigenvalues near the stationary point and
give us no trouble. In this manner we obtain Sagredo's answer,
Eq. (2.48), except that the negative eigenvalue carries a factor
of t with it; that is to say, we obtain Eq. (2.49).
III. THE VACUUM STRUCTURE OF GAUGE FIELD THEORIES 9
3.1 Old Stuff

This subsection is a telegraphic compendium of formulas from
gauge field theories. Its purpose is to establish notational con-
ventions and possibly to jog your memory. If you don't already
know the fundamentals of gauge field theory, you won't learn them
here. 10
Lie Algebras.
A representation of Lie a1gegra is a set of N
anti-Hermitian matrices, Ta , a = 1 ... N, obeying the equations
(3.1)
where the c's are the structure constants of some compact Lie group,
G. It is always possible to choose the T's such that Tr(TaT b ) is
proportional to cab, although the constant of proportionality may
depend on the representation. The Cartan inner product is defined
by
(3.2)
Thus this is proportional to the trace of the product of the

matrices.
So far I have not stated a convention that gives a scale to the
structure constants and thus to the T's. For SU(2), the case I will
abc abc
spend most time discussing, I will choose c to be equal to
Thus, for the isospinor representation,
where the 0's are the Pauli spin matrices. In this case,
Occasionally I will discuss SU(n) , in particular SU(3). In

this case I will choose the structure constants to agree with the
preceding convention for the SU(2) subgroup composed of unitary
unimodular transformations on two variables only. Thus, for SU(3),
Ta is -iA a /2, where the A's are Ge11-Mann's matrices.
828 S. COLEMAN
Gauge Fields. The gauge potentials are a set of vector fields,

Aa(x). It is convenient to define a matrix-valued vector field,
]l
A (x), by
]l A (3.5)
]l
where g is a constant called the gauge coupling constant. The

field-strength tensor, F (x), is defined by
]lV
F
]lV
= a] l AV - aVA] l + [A ,A ] .
]l v
(3.6)
Pure gauge field theory is defined by the Euclidean action,
S = 412
g
fd4X(F
]lV
,F
]lV
) . (3.7)
Sometimes I will write this in a shorthand form,
S 4!2 = f(F 2 ) (3.8)
Gauge Transformations. A gauge transformation is a function,

g(x), from Euclidean space into the gauge group, G. In equations,
where the A's are arbitrary functions. (Please do not confuse g(x)
with the coupling constant, g.) Under such a transformation,
(3.10)
and
(3.11)
Thus, S is gauge-invariant. If F vanishes, then A is a gauge-

]1\! ]l
transform of zero; that is to say,
A (3.12 )
]l
for some g(x).
Covariant Derivatives. The covariant derivative of the field
strength tensor is defined by
(3.13)
Equation (3.7) leads to the Euclidean equations of motion
D F = 0 (3.14 )
]l ]lv
Given a field ~ that gauge-transforms according to
~ + g(x)~ (3.15 )
then the covariant derivative of ~,
D~=a~+A~, (3.16)
]J ]J ]J
transforms in the same way.

3.2 The Winding Number
I propose to study Euclidean gauge field configurations of
finite action (not necessarily solutions of the equations of motion).
Why?
The naive answer, sometimes given in the literature,ll is that
configurations of infinite action are unimportant in the functional
.
1ntegra l '
,Slnce, f or suc h con f 19urat1ons,
. . e -s Prr 1S
. zero. Th ~s
. ~s
.
hlPong. In fact, it is configurations of finite action that are

unimportant; to be precise, they form a set of measure zero in
function space. This has nothing to do with the divergences of
quantum field theory; it is true even for the ordinary harmonic
oscillator. (For a proof, see Appendix C.) The only reason we are
interested in configurations of finite action is that we are in-
terested in doing semiclassical approximations, and a configuration
of infinite action does indeed give zero if it is used as the
center point of a Gaussian integral.
The convergence of the action integral is controlled by the
behavior of A for large r, where r is the radial variable in
]J
Euclidean four-space. To keep my arguments as simple as possible,
I will assume that, for large r, A can be expanded in an asymptotic
]J
series in inverse powers of r. (This assumption can be relaxed con-
siderably without altering the conclusions.~) Thus, for the action
to be finite, F must falloff faster than l/r2 as r goes to in-
lJ\l
finity; that is to say, F must be O(1/r 3 ). One's first thought
]J\l
is that this implies that A is O(1/r2), but this is wrong: vanishing
]J
F does not imply vanishing A , but merely that A is a gauge
~ ]J ]J
transform of zero. Thus A can be of the form
]J
830 s. COLEMAN
A (3.17)
].l
where g is a function from four-space to G of order one, that is to

say, a function of angular variables only.
Thus, with every finite-action field configuration there is
associated a group-e1ement-va1ued function of angular variables,
that is to say, a mapping of a three-dimensional hypersphere, S3,
into the gauge group, G. Of course, this assignment is not gauge
invariant. Under a gauge transformation, hex)
A -+ h A h- 1 + h d h- 1 (3.18)
].l].l ].l
Thus,
g -+ hg + O(1/r2) . (3.19)
If one could choose h to equal g-l at infinity, one could

transform g to one and eliminate it from Eq. (3.17). In general,
though, this is not possible. The reason is that h must be a con-
tinuous function not just on the hypersphere at infinity, but
throughout all four-space, that is to say, on a nested family of
hyperspheres going all the way from r equals zero to r equals in-
finity. In particular, at the origin, h must be a constant, inde-
pendent of angles. Thus, h at infinity can not be a general func-
tion on S3, but must be one that can be obtained by continuous
deformation from a constant function. Since any constant gauge
transformation can trivially be obtained by continuous deformation
from the identity transformation (all gauge groups are connected),
we might as well say that h at infinity must be obtainable from
h = 1 by a continuous deformation.
Given two mappings of one topological space into another, such
that one mapping is continuously deformable into another, mathe-
maticians say the two functions are "homotopic" or "in the same
homotopy class". What we have shown is that by a gauge transforma-
tion we can transform g(x) into any mapping homotopic to g(x), but
we can not transform it into a function in another homotopy class.
Thus, the gauge-invariant quantity associated with a finite-action
field configuration is not a mapping of S3 to G but a homotopy class
of such mappings. Our task is to find these homotopy classes for

physically interesting GIS.
To warm up for this task, let me consider a baby version of

the problem for which the geometry is somewhat easier to visualize.
I will work with the simplest of all gauge groups, U(l), the group
of complex numbers of unit modulus. Thus the gauge field theory
is ordinary electromagnetism. (However, I will still keep to the
notational conventions established in Sec. 3.1; in particular, A~
will be an imaginary quantity, i times the usual vector potential.)

Also, I will work not in Euclidean four-space but in Euclidean two-
space. I will still study fields obeying Eq. (3.17), although, of
course, in two-space this condition is not a consequence of finite-
ness of the action. Because we are working in two-space, we have,
instead of a hypersphere, S3, an ordinary circle, SI.
Now to work:
(1) G is the unit circle in the complex plane; thus, topolog-
ically, G is also SI, and we have to study homotopy classes of map-
pings of SI into SI. We will label the circle in space, the domain
of our functions, in the standard way, by an angle 6 ranging from
o to 21T.
(2) It will be useful to define some standard mappings from
SI to SI. One is the trivial mapping,
(3.20a)
Another is the identity mapping,

i6 (3.20b)
e
These are both part of a family of mappings,

iv6
e (3.20c)
where V is an integer (positive, negative, or zero). V is called

"the winding number", because it is the number of times we wind
around G when we go once around the circle at infinity in two-space.
(By convention, winding around minus once means winding around once
in the negative direction.)
832 S. COLEMAN
(3) Every mapping from Sl to Sl is homotopic to one of the

mappings (3.20c). We do not have the mathematical machinery to
prove this rigorously, but I hope I can made it plausible. Imagine
taking a rubber band and marking on it in ink a sequence of values
of 9 running from 0 to 2TI. We then wrap the band about a circle
representing G, such that each value of 9 lies above the point into
which it is mapped. (Figure 13 shows such a construction.) We can
continuously deform the band, first to eliminate any folds, like
the one on the top of the figure, and second to stretch the band so
it lies uniformly on the circle. In this way we obtain some g(V)(9).
(In the case shown, we obtain g(l).) Thus we can associate a wind-
ing number with every mapping. (Note that I have not yet shown that
this number is uniquely defined.)
(4) I will now show that the winding number defined above is
given by the integral formula
2TI
V = 2~ f d9gdg- l /d9 (3.21)
o
Firstly, by direct calculation, this gives the right answer for the
standard mappings, Eq. (3.20c). Secondly, this quantity is invariant
37T/2
o
Figure 13
under continuous deformations. To prove this assertion it suffices

to demonstrate invariance under infinitesimal deformations. A gen-
eral infinitesimal deformation is of the form
og = i(oA)g , (3.22)
where OA is some infinitesimal real function on the circle. Thus
O(gdg- 1 /d8) = -id(oA)/d8 , (3.23)
and the change in V vanishes upon integration. (We now know that
all of our standard mappings are in different homotopy classes and
that the winding number is uniquely defined.)
(5) If
g(8) (3.24a)
then
(3.24b)
The proof is simple: The winding number is unchanged by continuous

deformations. We can deform gl such that it is equal to one on the
upper half of the circle (0 ~ 8 ~rr) and g2 such that it is equal to
one on the lower half of the circle (rr ~ 8 ~ 2rr). The integrand in
Eq. (3.21) is then the sum of a part due to gl (vanishing on the
upper semicircle) and a part due to g2 (vanishing on the lower semi-
circle).
(6) Let us define
(3.25)
By Eqs. (3.17) and (3.21),

2rr
V = lim
r-xx>
J rd8r G
]l ]l
(3.26)
o
where r ]l is the radial unit vector. Thus, by Gauss's theorem,
(3.27)
Hence,
v=-i d x2 F
4rr ]lV ]lV
f
(3.28)
I will now return to four-space, and take G to be SU(2). As we

shall see, every argument will be a (mild) generalization of the
834 s. COLEMAN
arguments I have given for the baby problem.

(1) SU(2) is the group of unitary unimodular two-by-two
matrices. It is well known that any such matrix can be uniquely
written in the form -+ -+
g =a + ibo0 (3.29)
where a 2 + Ibl 2 = 1. Thus,topologically, SU(2) is S3, and we have

to study homotopy classes of mappings from S3 to S3.
(2) It will be useful to define some standard mappings from
S3 to S3. One is the trivial mapping,
g (0) (x) = 1 (3.30a)
Another is the identity mapping,
g (1) (x) = (x" + i-;.C;-) /r (3.30b)
These are both part of a family of mappings,
(3.30c)
where V is an integer, called the winding number. (It is also some-

times called the Pontryagin index.) It measures the number of times
the hypersphere at infinity is wrapped around G. (By convention,
we say the hypersphere is wrapped around G in a negative sense if a
right-handed triad of tangent vectors is mapped into a left-handed
triad.)
(3) Every mapping from S3 to S3 is homotopic to one of our
standard mappings (3. 30c). He do not have the mathematical machin-
ery to prove this assertion rigorously, but a plausibility argument
can be constructed just as in the baby problem, with hyperspheres
replacing circles. (If you have problems envisioning hyperspheres
wrapped around hyperspheres, just accept the assertion on faith.)
In this way we can associate a winding number with every mapping.
(Note that I have not yet shown that this number is uniquely defined.)
(4) Let us define
V = 48~2
1 fd8 1 d8 2 d8 3 E ijk( gOi
'\ g-1 , gd g- 1 gdkg- 1 )
j (3.31)
where 8 1 , 8 2 and 8 3 are three angles that parameterize S3. How

these angles are chosen is irrelevant to Eq. (3.31); the Jacobian
determinant that comes from changing the angles is canceled by the
Jacobian determinant from the E-symbol. Equation (3.31) is written
using the Cartan inner product, that is to say, in a representation-
independent way. Of course, for any particular representation of
SU(2), we can rewrite Eq. (3.31) in terms of traces; for example,
for the two dimensional representation, by Eq. (3.4),
V = - 1 Jd8 1 d8 2 d8 3TrE ijk"

24n2 gOi g-1"
gOjg-1 gOk
" g -1 (3.32)
I will show that this quantity is, firstly, a homotopy invari-

ant, and secondly, agrees with the winding number as defined for
our standard mappings. As before, a corollary of this proof will
be that all of our standard mappings are in different homotopy
classes and that the winding number is uniquely defined.
To show invariance under continuous deformations it suffices
to show invariance under infinitesimal deformations. For any Lie
group, a general infinitesimal transformation can be written as an
infinitesimal right multiplication:
a a
og = gOA (x)T - gOT . (3.33)
Under this transformation,
(3.34)
The three derivatives in Eq. (3.32) make equal contributions to OV;

thus,
(3.35)
If we use the identity,
(3.36)
this becomes
OV ~ Jd8 1d8 2d8 3E1]
k
Trdig-1djgdkoT , (3.37)
which vanishes upon integration by parts, because of the antisym-

metry of the E-symbol. This completes the proof of invariance under
836 S. COLEMAN
continuous deformations.
(5) Now to evaluate Eq. (3.32) for our standard mappings. The
task is easiest for g(l), for the integrand is here obviously a
constant, and we need evaluate it only at the north pole of the
unit hypersphere, x = 1, x. = O. At this point we might as well
It ~
choose 8. to equal x .. Thus, from Eq. (3.30b),

~ ~
d -1 .
(3.38)
g ig = -~oi '
and
(3.39 )
Since the area of a unit hyper sphere is 2n 2 , we obtain the desired

result, V = 1.
For the other standard mappings, the simplest way to proceed
is to observe that if
g (3.40a)
then
(3.40b)
The argument is the same as for the baby problem, with semihyper-
spheres replacing semicircles.
(6) Let us define
G (3.41)
lJ
A straightforward computation shows that
d G (3.42)
lJ lJ
The dual of an antisymmetric tensor (denoted by a tilde) is conven-

tionally defined by
(3.43)
(The factor of ~ is inserted in the definition so that F F. )

Equation (3.42) can thus be rewritten as
~ ~
d G (F ,F ):: (F,F) (3.44 )

lJ lJ lJv lJv
From the definition of FlJV'
GlJ = SlJv;\o(Av,F;\O- iA;\Ao) . (3.45)

This expression is useful in evaluating
fd4X (F,F) = ww '

Jd 3Sr G (3.46)
where d 3S is the element of area on a large hypersphere. The first

term in Eq. (3.45) is O(1/r4) and makes no contribution to the in-
tegral; the second term simply gives (up to a multiplicative con-
stant) the integral formula for the winding number, Eq. (3.31).
Thus we obtain f d4X (F,F) 32n 2 v . (3.47)
Summary and Generalizations. This has been a long analysis,

and you may have lost track of what we were doing, so let me sum-
marize the main results of this subsection: For a gauge field theory
based on the group SU(2), every field configuration of finite action
in four-dimensional Euclidean space has an integer associated with
it, the Pontryagin index or winding number, v. It is not possible
to continuously deform a configuration of one winding number into
one of a different winding number while maintaining the finiteness
of the action. He have two integral formulas for the winding number,
one in terms of a surface integral over a large sphere, Eq. (3.31),
and one in terms of a volume integral over all four-space, Eq. (3.47).
How much of this depends on the gauge group being SU(2)?
Firstly, if the gauge group is U(l), it is easy to see that every
mapping of S3 into U(l) is continuously deformable into the trivial
mapping (all of S3 mapped into a single point). Thus~ for an
Abelian gauge field theory~ there is no analog of the winding
number. Secondly, for a general simple Lie group, G, there is a
remarkable theorem due to Raoul Bott 14 that states that any continu-
ous mapping of S3 into G can be continuously deformed into a mapping
into an SU(2) subgroup of G. Thus~ everything we have discovered
for SU(2) is true for an arbitrary simple Lie group; in particular~
it is true for SU(n). I stress that "everything" means everything;

In particular, not a single numerical factor in the integral for-
mulas for the winding number needs alteration, so long as we choose
the normalization of the Cartan inner product appropriately (as we
838 S.COLEMAN
have). Finally, since a general compact Lie group is locally the

direct product of an Abelian group and a string of simple groups,
for a general gauge field theory, there is an independent winding
number for every simple factor group.
3.3 Many Vacua
We have learned a lot about classical gauge field theories;
now it is time to confront the quantum theory. In principle, the
Euclidean functional integral tells how to go from the classical
theory to the quantum theory. As I explained in Sec. 2, we can use
the functional integral to study the energy eigenstates of the the-
ory; also, by adding appropriate source terms to the Hamiltonian
(equivalently, to the Euclidean action) and then differentiating
with respect to the sources at the end of the computation, we can
study the expectation values of strings of operators, Euclidean
Green's functions. However, for gauge field theories, there is a
famous complication: to make the functional integral well-defined,
we must impose a gauge-fixing condition. IS
I will choose to work in axial gauge, A3 = O. I have several

reasons for this choice: (1) It is possible to show lli that every non-
singular gauge field configuration can be put in axial gauge by a
non-singular gauge transformation. It is by no means clear whether
this is true for covariant gauges, for example. (2) In axial gauge
the functional integral is directly equivalent to a canonical for-
mulation of the theory; 17 there is no need of the ghost terms that
occur in covariant gauges, or of the subsidiary conditions on the
space of states that are needed in such gauges as Ao = o. (3) t10st
of the treatment in the literature of the phenomena we are about to
discuss is in the gauge Ao = O. It's nice to show explicitly that
the answers don't depend on this gauge choice. (4) Although axial
gauge is terribly awkward for specific computations, once we have
obtained functional-integral expressions for quantities of interest,
we can use the standard Fadeev-Popov methods to transform these into
some more convenient gauge.
In field theory, we normally plunge directly into infinite

space. However, I will here study gauge field theory in a finite
box of three-volume V, with definite boundary conditions, which I
shall specify shortly. Just as in Sec. 2, I will also restrict the
theory to a finite range of Euclidean time, T, with appropriate
boundary conditions at initial and final times. Thus we are in-
tegrating over a box in Euclidean four-space, with boundary condi-
tions on the (three-dimensional) walls of the box. Of course, I
will eventually send both V and T to infinity. I again have reasons
for this choice: (1) Certainly nothing is lost by beginning in a
finite box; if the transition to infinite space goes smoothly, at
worst we will have wasted a little time. (2) In some theories, we
can gain information about the structure of the theory by seeing
how things depend on the boundary conditions imposed on the walls
of the box. For example, in a scalar field theory with spontaneous
symmetry breakdown, the expectation value of the scalar field in the
center of the box depends on the boundary conditions on the walls,
no matter how large the box; this is one of the easiest ways to see
that the theory has many vacua. (3) In the canonical quantization
of the theory, it is necessary to eliminate Ao from the action. To
do this, it is necessary to find Ao from a;Ao. In infinite space,
this problem has many solutions; this ambiguity is usually resolved
by applying ad hoc conditions on the behavior of Ao at infinity.
In a box with appropriate boundary conditions, this problem always
has a unique solution.
There are many possible types of boundary conditions we could
impose: we could fix some components of A , some components of F ,
]J ]J\i
some combinations of these, etc. A clue to a wise choice of bound-
ary conditions is given by the surface term in the expression for
the variation of the action. For example, for a free scalar field
theory,
(3.48)
Here, d 3 S is the element of surface area, n]J is the normal vector
to the surface, and the triple dots denote the usual volume integral
840 S. COLEMAN
of the Euler-Lagrange equations. From this expression we see that

one way to make the surface terms vanish is to fix the value of ~
on the walls of the box. Likewise, for a gauge field theory,
oS = \ fd 3 sn lJ F OAV + ... . (3.49)

g lJV
From this expression we see that one way to make the surface term
vanish is to fix the tangential components of A on the surface.
lJ
Note that there is no need to fix the normal component of A . be-
lJ'
cause F is antisymmetric, this makes no contribution to the sur-
lJV
face integral.
We are not totally free to choose the tangential components of
AlJ arbitrarily. Firstly, they must be chosen consistent with our
gauge condition, A3 = O. Secondly, because we want to do semi-
classical computations, we must choose our boundary conditions to
be consistent with finiteness of the action, as the box goes to
infinity. Equivalently, the boundary conditions must be consistent
with the box being filled with a field configuration of a definite
winding number. Furthermore, for fixed boundary conditions, this
winding number is fixed, for only the tangential components of AlJ
are needed to compute the normal component of G. [See Eq. (3.4l).J
lJ
Thus at least one relic of our boundary conditions remains no
matter how large the box: we can not put an arbitrary finite-action
field configuration in the box, but only one of a definite winding
number. It turns out that the winding number is the only relic of
the boundary conditions that survives as the box goes to infinity.
The hand-waving argument for this is that the winding number is the
only gauge-invariant quantity associated with the large-distance
behavior of the fields. If you do not find this argument convinc-
ing, you will find a more careful one in Appendix D.
Thus, for large boxes, we can forget about the boundary condi-
tions in the functional integral and simply integrate over all con-
figurations where the winding number, v, has some definite value, n.
I will denote the result of such an integration by F(V,T,n). In
equations,
-S
F(V,T,n) = N J [dA]e 0 . (3.50)
Vn
where [dA] denotes [dA l ][dA 2 ][dA 4 ]. Also, I have set-h to one; we
can always keep track of the powers of -tr by keeping track of the
powers of g, as explained in Sec. 1.
F(V,T,n) is a transition matrix element from some initial state
to some final state (determined by our boundary conditions). ~fuat
these states are will not be important to us. Hhat is important is

that for large times, Tl and T 2 ,
F(V,T l +T 2 ,n) I F(V,T l ,n l )F(V,T 2 ,n 2 ) (3.51)

nl+Dz=n
This follows from Eq. (3.47), the expression for the winding number
as the integral of a local density; this tells us that the way to
put total winding number n in a large box is to put winding number
n l in one part of the box and winding number n 2 in the remainder of
the box, with n = nl + n2 (Of course, such counting misses field
configurations with significant action density on the boundary be-
tween the two sub-boxes, for there is no reason for the winding-
number integral for each sub-box to be an integer for such config-
urations. However, we expect this to be a negligible surface ef-
fect for sufficiently large boxes.)
Pretty as it is, Eq. (3.51) is not what we would expect from a
transition-matrix element that has a contribution from only a single
energy eigenstate. Such an object would be a simple exponential,
and would obey a multiplicative composition law for large times, not
the convolutive composition law of Eq. (3.51). However, it is easy
enough to turn convolutions into multiplications. The technique is
called Fourier transformation:
\' ine
F(V,T,e) - L e F(V,T,n)
n
f
N [dA]e- S e iVe (3.52)
From Eq. (3.51),
(3.53)
842 S. COLEMAN
This is the correct composition law for a simple exponential. Thus

we identify F(V,T,8) as being (up to a normalization constant) the
.
expectatlon va 1 ue 0 f e -HT In
. an energy elgenstate,
. wh'lC h we d enote
by 18> and call the 8 vacuum.
F(V,T,8) cr <8Ie- HT I8>
N'f[dA]e- S e iV8 (3.54)
where N' is a new normalization constant.

Our analysis has been simple and straightforward (I hope), but
we have been led to a very unintuitive conclusion. Our original
gauge field theory seems to have split up into a family of discon-
nected sectors, labeled by the angle 8, each with its own vacuum.
Furthermore, in each of these sectors, the computational rules are
the same as those we would have naively written down if we had not
gone through any of this analysis, except that an extra term, pro-
portional to (F,F), has been added to the Lagrangian density.
Probably half the people who have played with gauge field theories
have thought, at one time or another, of adding such a term, and
they have discarded the possibility, because the added term is a
total divergence [see Eq. (3.44)] and thus has no effect on the
equations of motion and therefore "obviously" has no effect on the
physics of the theory. Of course, at this stage in our investiga-
tion, it is still possible that we have been fooling ourselves,
that the extra term indeed has no effect on the physics, and that
all the 8 vacua we think we have discovered are simply duplicates
of the same state. We shall eliminate this possibility immediately.
[I should remark that what we have done here closely parallels
the treatment of a periodic potential in Sec. 2.3, except the argu-
ments are somewhat more abstract and in a different order. The
winding number is something like the total change in x (the differ-
ence between the number of instantons and the number of anti-
instantons) in Sec. 2.3, and the 8 vacua are something like the 18>
eigenstates. The two big differences are that we found the analogs
of the 18> states without pausing to talk about the analogs of the
Ij> states, and that we did the Fourier transform that untangled
the energy spectrum before we saturated the functional integral
with instantons. The first difference is unimportant; if I had
wanted to, I could have added two extra paragraphs when I was talk-
ing about F(V,T,n) and discussed the analogs of the Ij> states.
(They're called n vacua.) As for the instantons, they are the sub-
ject of the next subsection.]
3.4 Instantons: Generalities
In the next subsection I shall explicitly construct instantons,
finite-action solutions of the Euclidean gauge-field equations with
V = 1. Most of the qualitative consequences of these solutions are
independent of their detailed structure and follow merely from the
fact of their existence. Therefore, in this subsection, I will
simply assume that instantons exist and draw some conclusions from
this assumption.
I will denote the action of an instanton by So' Because So
is finite, the instanton can not be invariant under spatial trans-
lations. Thus there exists at least a four-parameter family of in-
stanton solutions; I will call these parameters "the location of
the center of the instanton". The winding number is parity-odd.
Thus there must also exist at least a four-parameter family of solu-
tions with V = -1, the parity transforms of the instanton solutions,
which I will call anti-instantons. Just as in Sec. 2, we can build
approximate solutions consisting of n instantons and n anti-
instantons, with their centers at arbitrary widely-separated loca-
tions . These approximate solutions have V = n - n.
Again as in Sec. 2, we approximate Eq. (3.54) by summing over
all these configurations. Thus we obtain
<8Ie- HT I8> ex L (Ke- SO )n+i1(VT)n+TI e i(n-ll)8/ n !n!

n,n
-S
exp(2KVTe 0 cos 8) (3.55)
where K is a determinental factor, defined as in Sec. 2. Thus, the

844 S. COLEMAN
energy of a 8 vacuum is given by
E(8)/V = -2K cos 8 e- So (3.56)
Note that, as should be the case in a field theory, the different

vacua are distinguished not by different energies, but by different
energy densities. [Also note the similarity with the energy spec-
trum of a periodic potential, Eq. (2.45).]
He can go on and compute the expectation values of various
operators. A particularly easy (and particularly instructive)
computation is that of the expectation value of (F,F). By trans-
lational invariance,
<81 (F(x),F(x)) 18> = iT Jd 4 x<81(F,F) 18> (3.57)
Thus, by Eq. (3.47),

32n2I[dA]ve-SeiV8
<81 (F,F) 18>
VTI[dA]e- S e iV8
(3.58)
Hence there is no need to do a fresh summation over a dilute in-

stanton - anti-instanton gas, since we have just evaluated the
quantity in parentheses in Eq. (3.55). Thus in our approximation,
I
<8 (F, F) 18> = -64n 2 iKe -So sin 8 . (3.59 )
Some comments:
(1) The expectation value is independent of V and T, as it
should be.
(2) The expectation value is an imaginary number, again as it
should be. The reason is that
(3.60)
tihen we continue from Euclidean space to Minkowski space, F12 re-

mains F 12' but, just as x 4 becomes ixo' so does F 34 become iF 30
Thus, if we had obtained a real answer, we would have found that in

Minkowski space (the real world) a Hermitian operator would have
had an imaginary vacuum expectation value, a disaster.
(3) Both the vacuum energy density and the vacuum expectation
value depend non-trivially on 8. Thus the 8 vacua are indeed all
different from each other.
3.5 Instantons: Particulars
jd"x(F,F) [jd"X(F,F)jd"x(F,F)]~
~ Ijd"x(F,F)I , (3.61)
by the Schwartz inequality. Thus, for any winding number, we have

an absolute lower bound on the action,
Iv I
811'2
S ~ -2 (3.62)
g
Furthermore, equality is attained if and only if
F = F (3.63)
where the positive (negative) sign holds for positive (negative) v.

This inequality was first derived by Belavin, Polyakov,
Schwarz, and Tyupkin,9 who used it to search for instantons. Their
idea was to look for solutions of Eq. (3.63). If such solutions
exist, they are minima of the action for fixed winding number, and
thus stationary points of the action under local variations, that
is to say, solutions of the field equations. Furthermore, since
they have lower action than any other solutions of the same winding
number (if other solutions exist), they dominate the functional in-
tegral, and, for our purposes, are the only solutions we need worry
about. Finally, as a bonus, Eq. (3.63) is a first-order differen-
tial equation and considerably more tractable than the second-order
field equations.
Let us begin the search with v = 1. We know that any field
configuration with v = 1 can be gauge-transformed such that
A (3.64 )
11
where
(3.65)
Equation (3.64) is rotationally invariant, in the sense that the

effect of any four-dimensional rotation can be undone by an
846 S. COLEMAN
appropriate gauge transformation. This is a consequence of the

statement that a rotation is a continuous deformation and thus does
not change the winding number. There is also a short direct proof:
Under a general rotation
(1) (1) -1
g -+ gg h , (3.66)
where g and h are elements of SU(2) determined by the rotation.

[This is a standard formula; it is the usual way of demonstrating
the isomorphism between SO(4) and SU(2) 0 SU(2).J Thus,
A
jJ
-+ gA g -
. jJ
1 + 0 ( 1 /r 2 ) (3 . 6 7)
This, as promised, can be undone by a gauge transformation, indeed,

by a gauge transformation of the first kind, a constant gauge
transformation.
This suggests that we search for a solution of Eq. (3.63) that
is rotationally invariant in the same sense. That is to say, we
make the ansatz,
A (3.68)
jJ
where, to avoid a singularity, f must vanish at the origin. From

here on it's straightforward plug-in-and-crank, which I will spare
you. It turns out that we do indeed obtain a solution in this way,
if
f (3.69)
where p is an arbitrary constant, called "the size of the instan-

ton". The existence of solutions of arbitrary sizes is a necessary
consequence of the scale invariance of the classical field theory.
(This fact will occasion some embarrassment shortly.)
Once we have a solution to any field theory, we can obtain new
solutions by applying the invariances of the theory. In the case
at hand, these are generated by (1) scale transformations,
(2) rotations, (3) the four-parameter group of spatial transla-
tions, (4) the four-parameter group of special conformal transfor-
mations, and (5) gauge transformations. Scale transformations
simply change the size of the instanton; thus they just shift around
the members of our one-parameter family of solutions but generate

no new solutions. Rotations, as I have shown, can always be undone
by gauge transformations. Spatial translations generate genuinely
new solutions, and give us four more parameters, "the location of
the center of the instanton". Although I do not have time to
demonstrate it here, it turns out lS that special conformal transfor-
mations can be undone by gauge transformations and translations.
Gauge transformations, as usual, require special consideration.
It is easy to see that any non-trivial gauge transformation changes
(3.68): Because g(l)is a function of angles only, the radial compo-
nent of A , A ,vanishes. Thus, under a general non-singular gauge
jl r
transformation, g(x),
A -+ gA g -1 + gd g -1 = gd g -1 (3.70 )
r r r r
Hence, if the gauge transformation is not to change Ajl' g must be
independent of r. That is to say, its value everywhere must be its
value at the origin; g must be a constant gauge transformation.
But the only constant gauge transformation that leaves A unchanged
jl
is the identity. (Remember, the effect of a constant gauge trans-
formation is the same as that of a rotation.)
You might think that this discussion of gauge transformations
is irrelevant. After all, when we do the quantum theory, we must
work in a fixed gauge, such as axial gauge, and it is commonly said
that once we have fixed the gauge we have no freedom to make gauge
transformations. However, although commonly said, this is not
strictly true; all standard gauges still allow constant gauge
transformations. 19 This is as it should be. Constant gauge trans-
formations act like ordinary symmetries; they put particles into
multiplets (if there is no spontaneous symmetry breakdown), impose
selection rules on scattering processes, etc. Thus, in a sensible
formulation of the theory, they should remain as manifest symmetries
of the Hamiltonian. Hhether you accept this philosophy or not, the
fact remains that constant gauge transformation applied to an in-
stanton solution (transformed to obey the gauge conditions) will
generate a different solution still obeying the gauge conditions.
848 s. COLEMAN
Thus we have found an eight-parameter family of solutions, one

parameter from scale transformations, four from translations, and
three from constant gauge transformations.
Are there other solutions with unit winding number? Atiyah
and Ward~ state that there are none. I can not give their proof
here because I do not understand it. Nevertheless, mathematicians
I trust say that their argument is not only legitimate but bril-
liant, so let us assume they are right and continue.
Solutions of higher winding number (if they exist) are of no
interest to us. We have used approximate solutions consisting of n
widely separated objects (instantons or anti-instantons) to evalu-
ate the functional integral. These approximate solutions depend
on 8n parameters, 8 for each object. Now suppose there are exact
solutions that can be interpreted as n objects; that is to say,
they depend on 8n (or fewer) parameters and become our approximate
solutions when some of the parameters (the separations between the
objects) become large. In this case, all we learn by knowing these
exact solutions exist is that the dilute-gas approximation is better
than we think it is - but we already know that it is good enough for
our purposes. There might also be exact solutions that can not be
interpreted in this way. To have a definite example, let me suppose
there were a "binstanton", a brand-new solution of winding number
two. Then in evaluating the functional integral, we would have to
sum over a dilute gas of instantons, anti-instantons, binstantons,
and anti-binstantons. Thus, Eq. (3.56) would be replaced by
-S
E(8)/V -2Kcos8e 0
-S'
-2K'cos28e 0 (3.71)
where the primed quantities are the action and determinantal factor
for a binstanton. But S~ is twice So' so the new term is exponen-
tially small compared to the old one and should be neglected. 21
3.6 The Evaluation of the Determinant and an Infrared Embarrassment
We now know enough to go a long way towards explicitly evalu-
ating the right-hand side of Eq. (3.56).
(1) So is 8~2/g2.
(2) We have an eight-parameter family of solutions and thus
eight eigenmodes of eigenvalue zero in the small-vibration problem.
Thus K contains a factor of (1/~)8. or, equivalently l/g8. Every-
thing else in K is independent of~, and thus independent of g.
(3) He have already done the integral over instanton locations.
The integral over constant gauge transformations is an integral
over a compact group and thus gives only a constant numerical fac-
tor, the volume of SU(2). The integral over instanton sizes is
potentially troublesome, since p can be anywhere between zero and
infinity, so we will, for the moment, keep it as an explicit inte-
gral.
(4) Thus we obtain
J dp~
00
E(8) Iv = - cos 8 e
-8 ~21 g 2 g-8 f(pM) , (3.72)
o
where f is an unknown function and M is the arbitrary mass (more
properly, arbitrary inverse wavelength) that is needed to define
the renormalization prescription in a massless field theory. (I
have avoided mentioning renormalization until now, but renormaliza-
tion is essential in any computation that involves an infinite
number of eigenmodes, as does this one. In Sec. 5 I will give a
more detailed discussion of the ultraviolet divergences in deter-
minantal factors and their removal by the usual one-loop renormal-
ization counterterms.) The form of the integral is determined by
dimensional analysis; an energy density has dimensions of
II (length)" .
(5) However, M and g are not independent parameters. Renor-
malization-group analysis~ tells us that they must enter expressions
for observable quantities only in the combination
(3.73)
where Sl is a coefficient which can be computed from one-loop

850 s. COLEMAN
perturbation theory. In the case at hand, B1 is 11/12TI2.

(6) This fixes the form of f. Thus,
E(8)/V = _Acos8e-8TI2/g2 g-8r~ (PM)8TI 2B1[1+O(g2)],

o (3.74)
where A is a constant independent of g,p, and M.
(7) To determine A requires a lot of hard work,23so I'll stop
the calculation here. Even though we haven't been able to carry
things out to the end, it's remarkable how far we have been able to
go with so little effort.
No doubt you have noticed that the integral we have derived is
infrared divergent. The origin of the divergence is clear from the
derivation of the integral: the effective coupling constant (in the
sense of the renormalization group) becomes large for large ins tan-
tons, and this makes the integrand blow up. Thus the divergence is
an embarrassment but not a catastrophe. It would be a catastrophe
if we obtained a divergent answer in a regime in which we trusted
our approximations. This is not the situation here; the divergence
arises in the regime of large effective coupling constant, where
all small-coupling approximations are certainly wrong. Phrased
another way, the fact that the integrand has the wrong behavior for
large p is overshadowed by the fact that it is the wrong integrand.
Thus we are free to hope that strong-coupling effects (which we can
not at the moment compute) introduce some sort of effective infra-
red cutoff in the integrand. This hope might be wrong, but it is
not ruled out by anything we have done so far.
I admit that this argument is blatant hand-waving. However,
it is not some new hand-waving special to instanton calculations,
but the same old hand-waving that accompanies any discussion of the
large-scale behavior of non-Abelian gauge field theories. For
example, there is evidence that the observed hadrons are made of
weakly-coupled quarks. But if the quarks are weakly coupled, why
can't we knock them out of the hadron? Well, in a gauge field the-
ory the effective coupling constant grows at large distances,
etc., much hand-waving, infrared slavery and quark confinement.

Everything that we have done for SU(2) can be extended straight-
forwardly to SU(3). To begin with, an SU(2) instanton solution can
trivially be made into an SU(3) instanton solution; all that needs
to be done is to say that three of the gauge fields, those associ-
ated with an SU(2) subgroup, are of the form given, while the other
five vanish. It is believed that these exhaust the set of solu-
tions of Eq. (3.63) with unit winding number, although, unlike the
SU(2) case, there is, to my knowledge, no rigorous proof of this
statement. If this is indeed the case, there are only two minor
differences between the SU(3) computation and the SU(2) one:
(1) Instead of three parameters associated with constant gauge
transformations, we have seven. [One of the eight SU(3) generators
commutes with the SU(2) subgroup and does not change the solution.)
Thus the factor of g-8 in Eq. (3.74) is replaced by one of g-12.
(2) 6 1 has the proper value for an SU(3) gauge theory, 11/8n 2 .
IV. THE ABELIAN HIGGS MODEL IN 1 + 1 DIMENSIONS 2'+
In this section I will discuss a field theory in which instan-

ton effects drastically change the particle spectrum, the Abelian
Higgs model in two-dimensional space-time.
In any number of dimensions, this is the theory of a complex
scalar field with quartic self-interactions, minimally coupled to
an Abelian gauge field with gauge coupling constant e, called the
electric charge. In our notation, the theory is defined by the
Euclidean Lagrangian density,
(4.1)
where A is a positive number and ~2 may be either positive or nega-

tive. To this must be added renorma1ization counterterms; however,
renorma1ization will play no part in our computations, and, to keep
things as simple as possible, I will not distinguish between bare
and renorma1ized parameters.
852 s. COLEMAN
Perturbation theory tells us that for weak coupling the quali-

tativ~ properties of the theory depend critically on the sign of ~2:
(1) If ~2 is positive, the theory is simply the electrodynam-
ics of a charged scalar meson. The mass spectrum consists of the
charged meson, its antiparticle, and a massless vector meson, the
photon. The force between widely separated external charges is the
ordinary Coulomb force. These statements require some modification
in two dimensions. Firstly, because there are no transverse direc-
tions, there is no photon. Secondly, because the Coulomb force is
independent of distance, it is impossible to separate a meson and
an antimeson; in contemporary argot, the charged particles are con-
fined. The spectrum of the theory consists of a sequence of meson-
antimeson bound states, rather like the spectrum of positronium,
except that these states are all stable, since they can not decay
through the emission of (nonexistent) photons.
(2) If ~2 is negative, the Higgs phenomenon takes place. In
the ground state of the theory,
(4.2)
The particle spectrum consists of a massive neutral scalar meson

and a massive neutral vector meson. The force between widely
separated external charges falls off exponentially rapidly. These
statements require no modification in two dimensions.
In the remainder of this section, I will argue that the pre~
ceding sentence is a lie; contrary to the predictions of perturba-

tion theory, the qualitative properties of the model for negative
~2 are the same as those for positive ~2; the two-dimensional Abelian
Higgs model does not display the Higgs phenomenon. To be precise,
I will show that, for negative ~2, the theory admits instantons,
and, when the effects of these instantons are taken into account,
the long-range force between external charges is independent of their
separation. Also, I will be able to argue, from the behavior of the
long-range force, that the theory contains (confined) charged par-
ticles. There is a quantitative difference between positive and
negative ~2, though: For positive ~2, the strength of the long-range
force is independent of~; for negative ~2, the strength of the
long-range force is exponentially small in~, the mark of an in-
stanton effect.
Just as in Sec. 3, we must begin the analysis by classifying
classical field configurations of finite action. Of course, before
doing this, we must add a constant to the Lagrangian density so the
minimum of the action is zero. Thus we write
(4.3)
This is the sum of three positive terms. In order that the third
term not make a divergent contribution to the action, it is neces-
sary that I~I approach a as r goes to infinity. However, there is
no restriction on the phase of~. In equations,
lim ~ ( r , e) = g ( e)a , (4.4)

r~
where g is a complex number of unit modulus, an element of U(l).

In order that the second term not make a divergent contribution to
the action, it is necessary that
A (4.5)
~
(Remember, in our conventions, A is an imaginary field.) The first

~
term now automatically makes a finite contribution to the action.
The lovely thing about Eq. (4.5) is that it is identical to
Eq. (3.17); that is to say, the problem of classifying finite-
action configurations is the baby problem of Sec. 3.2. Thus the
finite-action configurations are characterized by an integer, v,
the winding number, just as they are for four-dimensional gauge
field theories. By Eq. (3.28), the integral expression for the
winding number is
(4.6)
Equivalently,
v =~
27T
fA~ dx~' (4.7)
where the integral is over the circle at infinity.

854 s. COLEMAN
Although I won't bother to explicitly display them here, it

turns out that the Euclidean field equations have solutions with
unit winding number, instantons, again just like four-dimensional
gauge theories.~ The only relevant difference, for our purposes,
is that the Higgs model is not scale invariant; thus the instantons
have a fixed size and the problems associated with integrating over
scale transformations don't arise. Otherwise, though, everything
is much the same as it was before, and we can copy step-by-step our
earlier analysis and uncover the vacuum structure of the theory.
Thus, just as before, we have a family of 8-vacua, with energy
densities given by
E(8) /L = -2Ke -Socos 8 . (4.8)
Here L is the volume of (one-dimensional) space, So is the action

of an instanton, and K is a determinanta1 factor. Also, by copying
the derivation of Eq. (3.59), we find that
<8 I F 18> = 87fKe-So sin 8 (4.9)

~v ~v
As before, this has the right reality properties; when we continue

to Minkowski space, we pick up a factor of i that cancels the fac-
tor of i in our definition of A~. We see from this equation that
the 8-vacua are characterized by a constant expectation value of
the electric field F 01 ' In two dimensions, unlike four, such a con-
stant "background field" is not in conflict with Lorentz invariance.'4.
Now that we understand the vacuum structure, let's compute the
force between widely separated external charges. To be more precise,
let us introduce into the system two static charges of equal magni-
tude, q, and opposite sign, separated by a distance L', and let us
compute (for large L') ~, the change in the energy of a 8-vacuum
caused by these charges. The standard method of computing ~ uses
fA~ dX~)
Wilson's loop integra1,27
W = exp (- ; , (4.10)
where the integration is over the rectangular path shown in Fig. 14.
According to Wilson, the vacuum energy shift is given by
TT'
1
Figure 14
- lim T~ n<SIWIS> . (4.11)

T'-->oo
In our case,
![dA] [d1jJ*] [dl/l]\.Je-Se iVS
<slwls>
J[dA] [dl/l*] [dl/l] e -S e iVS (4.12)
and our task is to compute these two functional integrals in our

standard dilute-gas approximation, for large L' and T' (and, of
course, for even larger Land T, the spatial and temporal extent of
the universe). In Eq. (4.8) we have already calculated the denomi-
nator. To calculate the numerator, let us divide the sum over in-
stantons and anti-instantons into two independent sums: one over
objects lying inside the loop and one over objects lying outside
the loop. By this division we neglect contributions coming from
configurations in which instantons and anti-instantons overlap the
loop, but, for large L, T, L', and T', this is a very small portion
of the available configurations and can reasonably be neglected.
(Of course, if our calculation gives zero for its answer, then these
configurations will be the most important ones and we will have to
go back and compute them.) The functional integrand splits neatly
into the product of an "outside" term and an "inside" term:
S = soutside + Sinside, V = voutside + vinside, while
856 s. COLEMAN
. inside
l>l = exp(Z1Tlq \! / e) . (4.13)
Thus, for the outside objects, we have the same sum as for the
denominator, except that the available volume of Euclidean two-
space is not LT but LT -L'T'. For the inside objects, we also have
the same sum, except that the available volume is L'T', and S is
replaced by S + Znq/ e .
Thus,
-S
.Q,n<SIHls> ZKe 0 [(LT -L'T')cos S
+ L'T'cos(S +Znq/e)
- LT cos S] , (4.14)
where the first term comes from the outside sum, the second from
the inside sum, and the third from the denominator. Hence,
t:, = ZL'Ke-SO[cosS-cos(S+Znq/e)] . (4.15)
This is proportional to L', the separation between the external

charges; thus there is a constant force between external charges
at large separation. As announced, there is no quantitative dif-
ference between positive and negative ~2. However, there is a
qualitative difference. For positive ~2, the strength of the force
is proportional to q2 for small~; for negative ~2, it is exponen-
tially small in1f. (Remember, if we had not chosen our units so ~
was one, So would have been So/~.)

There is a simple physical interpretation of this result. For
small S and small q/e,
t:, = L'Ke- SO [(S +q/e)2 - s2] , (4.16 )
E(S) = LKe -S oS 2 + constant , (4.17)

and,
(4.18)
These expressions have an obvious interpretation: In a S-

vacuum, there is a background electric field, and an energy density
proportional to the square of this field. Because we are in one
spatial dimension, the external charges act like condenser plates

in three dimensions; they induce a constant field proportional to
their charge in the region between them, which is added to the pre-
existing background field. Thus the energy shift is the separation
multiplied by the difference of the energy density of the new field
and that of the old. Equation (4.14) is just this trivial picture
complicated by non-linear terms in the expression for the energy
density as a function of the field.
One aspect of these non-linear complications is of physical
import: Eq. (4.14) is periodic in q with period e. This is explic-
able if the theory contains charged particles of charge e. If this
is the case, there is a process that can change the charge on our
condenser plates by e: a particle-antiparticle pair can material-
ize in the region between the plates, and the particle can fly to
one plate and the antiparticle to the other. This process will
occur whenever it is energetically favorable. For sufficiently
large L', this is equivalent to saying that it will occur whenever
it lowers the energy density, because the energetic cost of making
a pair is independent of L', and the energetic gain of lowering the
energy density is proportional to L'. Thus q's that are equal
modulo e lead to identical physics; no matter which one you start
out with, pairs are made until the charge on the plates reaches its
optimum value, the one that gives minimum energy density.
What if we were to do a parallel computation in a four-
dimensional gauge field theory, with non-Abelian external charges?
Would we also obtain a force independent of separation? Alas, we
would not. There is an L' in Eq. (4.15) because there is an L'T'
in Eq. (4.14), that is to say, because even an instanton deep within
the loop has a non-negligible effect on the loop integral. This is
precisely what does not happen in four dimensions. At large dis-
tances from an instanton, A is gd g-l, plus terms that falloff
~ ~
far too rapidly to affect the loop integral. However, the loop in-
tegral is gauge-invariant, and we can always gauge-transform g such
858 S. COLEMAN
that it is constant everywhere except within a small cone emerging

from the instanton perpendicular to the plane of the loop. l.Jhat-
ever confines quarks, it's not instantons.
V. 'T HOOFT'S SOLUTION OF THE U(l) PROBLEM
5.1 The Mystery of the Missing Meson

The U(l) problem is an apparent contradiction between two
pieces of accepted wisdom. One is wisdom of the '70's, that hadronic
physics is quantum chromodynamics. The other is wisdom of the '60's,
that hadronic physics is approximately invariant under chiral
SU(2) SU(2). Let me remind you of the meaning of these two prop-
ositions.
Quantum chromodynamics is a field theory whose dynamical vari-
ables are an octet of SU(3) gauge fields and a family of SU(3) trip-
let Dirac bispinor fields, called quarks. In Minkowski space, the
Lagrangian density is
..., 1 l1V \' - . 11
fJ....=-4 2 (F , F ) +L1/Jf(lD Y -m f )1/Jf' (5.1)
g l1 v f 11
where f, called the flavor index, labels the various triplets. The
usual exact and approximate symmetries of hadron physics [charge,
isospin, Gell-Hann' s sU(3) , etc.] act only on the flavor indices;
all physical hadrons are supposed to be singlets under the gauge
group. (This last statement is sometimes called quark confinement;
it is still far from proved, although there are some suggestive
arguments.) 1/J 1 and 1/J 2 form an isodoublet, the non-strange quarks;
1/J 3 is the strange quark; 1/J 4 is the charmed quark; there mayor may
not be additional flavors.
Chiral SU(2) SU(2) is the group generated by the strangeness-
conserving weak-interaction currents and their parity transforms.
Its diagonal subgroup is conventional isospin. This group is very
close to being an exact symmetry of the strong interactions; it is
a much better symmetry than SU(3) and roughly as good a symmetry as
isospin. However, were this symmetry to be exact, only the isospin
subgroup would be a manifest symmetry; the remainder of the group

would be a Nambu-Goldstone symmetry, with three massless Goldstone
bosons, the pions. The smallness of the pion mass (on a hadronic
mass scale) is a measure of the goodness of the symmetry. This is
the picture that stands in back of all the stunningly successful
soft-pion computations of the mid '60's.
Now for the apparent contradiction: In quantum chromodynamics,
the limit of perfect SU(2) 8 SU(2) symmetry is the limit in which
the non-strange quarks are massless. In this limit, the Lagrangian
(4.1) obviously has a further chiral U(l) symmetry; it is invariant
under
,1. -+ -ia.Y 5 ,1.
(f=1,2) (5.2)
'Vf e 'Vf '
where a. is a real number. The associated conserved current is
(5.3)
I emphasize that the appearance of this additional chiral symmetry

is very special to quantum chromodynamics; for example, the a model
has no such additional symmetry in the chiral limit.
Now, either this additional symmetry is manifest or it is spon-
taneously broken. If it were manifest, all non-massless hadrons
would occur in parity doublets. This is not the case; thus it must
be spontaneously broken. But if it is spontaneously broken,
Goldstone's theorem tells us there must be an associated isoscalar
pseudoscalar Goldstone boson. This is the U(l) problem: What
happened to the fourth Goldstone boson?
One's first thought is that the missing meson is the eta, but
this is wrong. The chiral U(l) symmetry is broken by the same mass
tern that breaks chiral SU(2) 8 SU(2), and thus the fourth Goldstone
boson should have roughly the same mass as the pions. The eta is
far too heavy. This can be made more precise: Using conventional
soft-pion methods, ~veinberg28 has shown that a U(l) Goldstone boson
must have a mass less than I3 IDn The eta grossly disobeys this
inequality. Also, if we consider the approximation in which the
860 s. COLEMAN
strange quark mass also vanishes, and in which we have perfect
chiral SU(3) 3 SU(3) symmetry, the eta takes its place with the
pions in an octet of Goldstone bosons. But in this limit we still
have an additional U(l) symmetry and we still have a missing meson.
[This should be all that I need to say about the eta. However,
there is some confusion abroad on this point, and thus I emphasize
that there is no connection between the eta and the U(lJ problem.
The eta is a red herring; it is just another hadron; it is no more
a relic of a U(l) Goldstone boson than is the N**.]
It may seem that I have posed an insoluble problem; this is
because I have lied to you. In fact, j~ is not a conserved current;
it is afflicted with the famous Adler-Bell-Jackiw anomaly.~ In the
limit of N massless quarks,
fl.5 N fl\,.lAo
a Jfl = 32~2 (FflV,F Ao ) (5.4)
(Note the similarity between the right-hand side of this equation

and the Pontryagin density. This will be important to us later.)
You might think that this is the end of the story; if the cur-
rent is not conserved, there is no U(l) symmetry to worry about.
Alas, life is not so simple. In Sec. 3, we showed that the Euclidean
counterpart of the right-hand side of Eq. (5.4) could be written as
the divergence of a (gauge-variant) function of
,.. and Ffl\,.l It is~,
easy to see that the same construction works in Minkowski space.

Thus, if we define
(5.5)
this current is gauge-variant but conserved.

If we work in a covariant gauge (and why shouldn't we?), the
added term commutes with the quark fields at equal times. Thus we
can derive, for Green's functions made of one J5 and a string of
fl
gauge-invariant quark multilinears, chiral U(l) Ward identities of
the usual form. And since these are of the usual form, they lead
to the usual conclusion: Chiral U(l) is a symmetry; either Green's
functions made of quark multilinears alone are D(l) symmetric, or

there are Goldstone poles in Green's functions for one J5 and a
]J
string of quark multilinears.
Is there no way out? Well, there is one. The Hilbert space
of a gauge field theory quantitized in a covariant gauge is notori-
ously full of negative-norm timelike photons and similar gauge
phantoms, states that never couple to gauge-invariant operators.
Could it be that the Goldstone boson is such a phantom? No, this
is not possible; the formulation of the question is wrong. If the
Goldstone boson does not couple at all to gauge-invariant operators,
it con't produce a pole in a Green's function for one J5 and a
]J
string of gauge-invariant operators.
The proper formulation of the question was found by Kogut and
Susskind,~who had the bright idea of looking at the Schwinger
model, massless spinor electrodynamics, in 1 +1 dimensions in a co-
variant gauge. The Schwinger model is an exactly soluble theory
that has properties very close to those we have been discussing.
In particular, there is a gauge-invariant axial current with an
anomalous divergence and a gauge-variant conserved axial current,
and, most important, there is chiral symmetry breakdown without
Goldstone poles in gauge-invariant Green's functions. What Kogut
and Susskind found in the covariant-gauge Schwinger model were two
free massless fields, <P + and <P _. <P+ creates quanta of positive
norm and has the usual propagator; <P creates quanta of negative
-
norm and has minus the usual propagator. (Remember, a covariant
gauge is full of negative-norm states from the very beginning.)
All gauge-invariant quantities couple to the sum of these fields,
+ + <P_; this has zero propagator and produces no singularities.
Thus gauge-invariant Green's functions are free of Goldstone poles.
However, the gauge-variant conserved current couples to the gradient
of the difference, 3 (<p+ - <P ). Thus, when one considers a Green's
]J -
function for one gauge-variant current and a string of gauge-
invariant fields, the relative minus sign in the coupling cancels
862 S. COLEMAN
the relative minus sign in the propagators, and Goldstone poles ap-
pear where they should. This set-up is called a Goldstone dipole.
(The terminology is a bit misleading, because there are only single
poles in Green's functions, but I'll stick with it anyway.)
Thus according to Kogut and Susskind, the proper formulation
of our question is, is the D(l) symmetry of quantum chromodynamics
spontaneously broken via a Goldstone dipole? You might think that
this is a question that could be asked seriously only by a field
theorist driven mad by spending too many years in too few dimensions.
Nevertheless, as 't Hooft 9 brilliantly showed, the answer is yes.
The remainder of this section is an explanation of his computation.
5.2 Preliminaries: Euclidean Fermi Fields
Before we can treat quantum chromo dynamics by functional inte-
gration, we must know how to integrate over Euclidean Fermi fields.
This section is a description of the theory of such integration,
with all mathematical fine points ruthlessly suppressed.~ I will
develop the theory by defining Fermi integration as a "natural"
generalization of Bose integration. At the end, I will justify my
definitions by showing that they lead to formulas equivalent to
those obtained by conventional canonical quantization.
Let us begin by defining our integration variables. For Bose
theories, we integrate over c-number Euclidean fields. These are
objects that commute with each other at arbitrary separations; they
can be thought of as the classical (vanishing~) limit of quantum
Bose fields. This suggests that the proper variables for a Fermi
theory should be classical Fermi fields, objects which anticommute
with each other at arbitrary separations. Thus, for example, for
the theory of a single Dirac field, we would expect our integration
variables to be two Euclidean bispinors, ~ and ~, obeying
{~(x),~(y)} = {~(x) ,~(y)} = {~(x) ,~(y)} = 0 , (5.6)
for all Euclidean points x and y.

The last of these relations is crucial, for it implies that ~
can not be in any sense the adjoint of ~ times some matrix. For if
this were so, the last relation (multiplied by the inverse matrix)
would state that the sum of two positive semi-definite objects,
~~+ and ~+~, was zero. This would only be possible if ~ vanished,
not a happy situation for a prospective integration variable. Thus,
if we are to have any hope of founding a sensible integration the-
ory, we must treat ~ and ~ as totally independent variables.
This independence is the main novelty of Euclidean Fermi
fields; the rest of the construction is straightforward. We define
the Euclidean Y-matrices to be four Hermitian matrices obeying
{Y,Y}=20. (5.7)
)J v )JV
We use these to define the 0(4) transformation law for ~ in the

usual way, and define \jJ to transform like the adjoint of ~. 1\Te
define Ys ' a Hermitian matrix, by
(5.8)
Thus, ~ is a scalar, ~Ys~ a pseudosca1ar, \jJY)J~ a vector, etc.

The Euclidean action for a free Dirac field is
S = - Id"XijJ(iCl Y -
)J )J
im)~ . (5.9)
The minus sign is pure convention; we could always absorb it into

~ if we wanted to. (Remember, we are free to transform ~ without
touching ~.) The i in front of the mass term is not conventional.
It is there to insure that the Euclidean propagator is proportional
to (p+im)/(p2+ m2); i f i t were not for the i, we would have tachyon
poles. If m vanishes, Eq. (5.9) is invariant under chira1 trans-
formations,
,," -iaY - - -iaY
~ + e s~, ~ + ~e s. (5.10)
The quark part of the Euclidean action for quantum chromodynamics

is obtained from Eq. (5.9) by replacing ordinary derivatives by
covariant derivatives.
So much for the integrand; now for the integration. For Bose
fields, we defined functional integration as iterated integration
over ordinary numbers. Therefore, let us begin by defining
864 S. COLEMAN
integration for a function of a single anticommuting quantity, a.

(Of course, for a single quantity, the anticommutation algebra
degenerates to a single equation, a 2 = 0.)
He want to define
fdaf (a) , (5.ll)
for an arbitrary function, f. He want this to have the usual

linearity property: the integral of a linear combination of two
functions should be the linear combination of the integrals. In
addition, we would like the integral to be translation-invariant
fdaf(a+b) = fdaf(a) (5.12)
where b is an arbitrary anticommuting quantity. I will now show

that these conditions determine the integral, up to a normalization
factor.
The reason is that there are only two linearly independent
functions of a, 1 and a; all higher powers vanish. vIe will choose
our normalization such that
fda a = 1 . (5.13)
From this, and Eq. (5.12),

fda 1 = 0 . (5.14)
For functions of many anticommuting variables, we define mul-

tiple integrals as iterated single integrals. Thus, for example,
a complete integration table for the four linearly independent func-
tions of two anticornrnuting variables, a and a, is
(5.15)
As an application of this table, I will evaluate
fdada- e Aaa = fdada(l + Aaa)

(5.16)
We can now define integration over Fermi fields exactly as we

defined integration over Bose fields in Sec. 2. vle introduce two
arbitrary complete orthonormal sets of c-number functions, ~r and
(5.17)
\~e expand the Fermi fields in terms of these functions,
~ ==La ~
r r r
(5.18)
and define
[d~] [d~] IT da da (5.19)
r r r
As an apDlication let me evaluate
(5.20a)
where
S == - Jd 4 x~A~
-
, (5.20b)
and A is some linear operator, possibly depending on external c-

number fields. For simplicity, let me assume that A commutes with
A+. (This is the case for a quark in an external gauge field.)
Then we can choose the ~ 's to be the eigenfunctions of A,
r
A~ (5.21)
r
and we can choose ~ to be ,1,+. Thus
r 'I'r
S - rLAr a r a
r
(5.22)
and
J[d~] [d\jJ] e -S ITA
r r
det A (5.23)
Note that this is the inverse of the answer we would have obtained
had we done the identical integral with ~ and ~ complex Bose fields.
I will now show that Eq. (5.22) is the correct answer, that it
is identical to the normal field-theoretic expression for the
vacuum-to-vacuum transition amplitude in a theory of a quantized
Dirac field interacting with external c-number fields. In this
theory, this amplitude is the sum of all Feynman graphs with no ex-
ternal Fermi lines. This in turn is the exponential of the sum of
866 s. COLEMAN
all connected (that is to say, one-loop) graphs. Now, if ~ were a

Bose field, we know that the amplitude would be the inverse deter-
minant, because we trust functional integration for Bose fields.
But the only effect of replacing bosons by fermions is to multiply
the one-loop graphs by minus one. This inverts the exponential of
the one-loop graphs, that is to say, it turns the inverse deter-
minant into the determinant.
In any theory in which the Fermi fields enter the action at
most bilinearly, we can always integrate over the Fermi fields,
using Eq. (5.23), before we integrate over the Bose fields. In
diagrammatic language, we can always sum the Fermi loops before we
integrate over virtual bosons. Thus, because our definition of
Fermi integration gives the right answer for a Dirac field in an
external c-number fields, it also gives the right answer for a
Dirac field interacting with a quantum Bose field. In particular,
it gives the right answer for quantum chromodynamics.
5.3 Preliminaries: Chiral Ward Identities
In this section is a discussion of the chiral Ward identities
for a theory of a set of quantum Dirac fields interacting with c-
number gauge fields. In the sequel, we shall use these identities
in several different cases; thus it's useful to have them written
down in their most general form, at hand when we need them.
Let ~ be a set of Euclidean Dirac fields, assembled into a big
vector, which transforms according to some representation of SU(n),
not necessarily irreducible, generated by a set of matrices, Ta .
Let us define the constant C by
Tr Ta Tb = _ C Gab . (5.24)
Thus, for example, for a set of N fields each transforming according

to the n-dimensional representation of SU(n) ,
C = N/2 . (5.25)
We wish to study the theory of these fields interacting with

given c-number gauge fields,
S = - i fd"x\i!(Y D -M)1/I , (5.26)

]J ]J
where DV is the covariant derivative defined by Eq. (3.16), and M

is the mass matrix for the Dirac fields, assumed to be SU(n)-invari-
(r)
ant. Let cP , r = 1 ... m, be a set of local multilinear functions
of the Dirac fields. The Euclidean Green's functions for these
objects are defined by
f[d1/l][d\i!]e- S (5.27)
where I have inserted the superscript A to remind you that we are

working in an external gauge field.
Now let us perform an infinitesimal change of variables in the
numerator of Eq. (5.27),
(5.28a)
where oa is an infinitesimal function of Euclidean space. Since

the CP's are functions of the Dirac fields, they will change under
the change of variables; we define 3cp(r)/3a by
(5.28b)
Thus, for example, 3~1/I/3a is -2i\i!Y s1/I. A change of variables

does not change the integral; thus, taking the variational deriva-
tive with respect to oa, we find
3]J<jS(y)cp(1) (xl) ... cp(m)(x A

]J m
- (1) (m) A
+ <1/IMY s1/I(y)CP (xl)CP (xm
+ o(4)(y_x )<3CP(1)(x )/3a ... cp(m)(x

1 1 m
A
o ,
(5.29)
where jS is 1/IY Y 1/1.
]J ]J 5
These are, of course, just the Euclidean version of the Hard
identities we would have obtained in Minkowski space by studying
868 S. COLEMAN
the divergence of j 5, and, of course, they are wrong, for they take
"f.!
no account of the Adler-Bell-Jackiw anomaly. I don't have the time
here to recapitulate the theory of the anomaly, and I will simply
state the correct version of Eq. (5.29): The zero on the right-hand
side is replaced by
i C ( -) (1) (m) A
- 8n2 F(y),F(y) < (Xl)' . (Xm . (5.30)
We can obtain a very useful equation by integrating the cor-

rected Ward identity over y. The first term on the left vanishes
by integration by parts; the theory contains no massless particles
that could give a non-vanishing surface term. Also, on the right
we can use
(3.47)
Thus we obtain
. (1) (m) A
= - 4lCV< (x) ... (x (5.31)
I m
Now all our artillery is at the ready; we can begin our assault
on quantum chromodynamics.
5.4 QCD (Baby Version)
I will begin by analyzing a baby version of quantum chromody-
namics, in which the gauge group is SU(2), and in which there is
only a single isodoublet quark, of mass zero. In equations,
S = f ,,[l .-
dXL4g2(F,F)-l\)JD"f.!Y"f.!\)J
J (5.32 )
After we have worked out the baby theory, we will go on to the real
thing.
Most of the analysis of Sec. 3 is essentially unaltered by the
presence of a quark. In particular, all of our old instanton solu-
tions are still solutions of the Euclidean equations of motion
(with the quark fields set equal to zero). Thus we still have all
the 8-vacua, and formulas like
-S (3.56)
E(8)/V = - 2Kcos 8 e O '
and
<8j(F,ihj8> = -647T 2 iKe- SO sin8, (3.59)
remain unaltered. The only effect of the quarks is to insert into

the definition of K a term proportional to
[U~J
det i~J (5.33)
where A is the field of an instanton.

II
This is a trifling alteration, but it is a tremendous trifle,
for, as we shall see, i~ has a vanishing eigenvalue. Thus the
determinant vanishes, as does E(8)/V and <8j(F,F) j8>!
The vanishing eigenvalue can be demonstrated either by a short
explicit computation or by a long indirect argument. I will choose
the second method. Despite what you might think, this is not a
perverse choice. (Well, not totally perverse.) The indirect argu-
ment will have some byproducts that will be very useful to us later.
For simplicity, I will assume (falsely) that i~ has a purely
discrete spectrum, ~
i~ \jJ = A \jJ (5 .34 )
r r r
Because i~ is Hermitian, all the A'S are real. Because Y s anti-
commutes with Yll '
i~ Y \jJ = - A Y \jJ (5.35 )
5 r r s r
Thus nonvanishing eigenvalues always occur in pairs of opposite
sign. Eigenfunctions of vanishing eigenvalue, on the other hand,
can always be chosen to be eigenfunctions of Y s '
(A 0) (5.36 )
r
Because Y~ = 1, Xr l. I will denote the number of eigenfunctions

of these two types by n.
I will now prove the remarkable sum rule,33
n -n =\!. (5.37)
+
Thus, not only is there a zero eigenvalue in the field of an in-
stanton, there is a zero eigenvalue in any gauge field of non-zero
winding number, whether or not it is a solution of the Euclidean
equations of motion.
870 s. COLEMAN
The proof rests on the chiral Ward identities for the quantum
theory of a massive quark interacting with an external gauge field.
(5.38)
If we take the case of no 's, Eq. (5.31) becomes
- 2iv 2<Jd'+ylfJ m'Ys1jJ>A
2f[d1jJ][dlfJ]e- S Jd 4 y lfJ m 'Y 5 1jJ

(5.39)
J[d1jJ][ dijJ] e -S
(Remember, in the case at hand, C = ~.) To evaluate the functional
integrals, we need the eigenfunctions and eigenvalues of i(fl- m).
The eigenfunctions are those of fl, and the eigenvalues are simply
shifted by -im,
i(~ - m)1jJ (A - im)1jJ . (5.40)
r r r
If we expand the fields in the 1jJ IS, the functional integrals
r
become trivial, and we
(5.41)
Because eigenfunctions of a Hermitian operator with different eigen-

values are orthogonal,
Jd 4 y1jJ;'Ys1jJr 0 if A F 0 , (5.42)
r
while
Jd 4 y1jJ;'Ys1jJr = Xr if A 0 (5.43)
r
Thus,
- 2iv 2i(n+ - n ) (5.44)
This is the desired result.

It turns out that the instanton obeys the sum rule by having
one eigenfunction of vanishing eigenvalue with X = -1 and none with
X = +1. (This also can be seen indirectly, without dirtying one's
hands with explicit computations; see Appendix E.) We'll never
need the explicit form of the eigenfunction, but, just for complete-
ness, I'll write it down here. For an instanton with center at X

and size p,
(5.45)
where u is a constant spinor. Likewise, for an anti-instanton,

there is one eigenfunction of vanishing eigenvalue with X= +1, the
parity transform of Eq. (5.45). For n widely separated instantons
and anti-instantons, there are n such eigenfunctions, one centered
about each object. (More properly, I should say that there are n
approximate eigenfunctions with approximately vanishing eigen-
values, but, for the dilute-gas approximation, the qualifications
are irrelevant.)
What is important for our purposes is that the sum rule implies
that any field configuration with nonvanishing winding number has at
least one eigenfunction of vanishing eigenvalue and thus a vanish-
ing Fermi determinant. Thus, not just in the dilute gas approxima-
tion, but to all orders in the semiclassical expansion, all the 8
vacua have the same energy and they all have a vanishing expectation
value for (F,P).
A phenomenon this general must have a deep cause. We can dis-
cover this cause if we consider the chira1 Ward identities for
vanishing quark mass. There is a technical obstacle to this; for
vanishing quark mass, the denominator in Eq. (5.27) vanishes, at
least for fields with V F O. This is easily surmounted; we define
denominator-free Green's functions,
cP
(1)
(xl)"'
A -
=
f [dl/J][dl/J]e
- -S cP (1) (Xl)'" (5.46)
By the same reasoning as before, these obey the Ward identities,
[~+
oCt
2iV]CP (x ) .. A = 0 ,
I I
(5.47)
i.e., Eq. (5.31) without the mass term. The Green's functions of
our baby version of chromodynamics are given by
f[dA]e -Sg e iV8 cp (1) (xl)" .A

<81 cP (1) (xl) . 18> = ----~---,.--:--~-
f[dA]e -Sg e iV8 1A (5.48)
872 s. COLEMAN
where S is the gauge-field part of the action. By Eq. (5.47),

g
3
[ 3a + 2aeJ<81 (1) (xl) ... 18>
3]
= 0 . (5.49)
Thus, the effect of a chiral U(l) transformation can be undone by

a change of 8. That is to say, chiral U(l) transformations turn
one 8 vacuum into another; chiral U(l) symmetry is spontaneously
broken, and the 8 vacua are the many vacua that appear when a sym-
metry suffers spontaneous breakdown. This is startling; after all,
when we first met the 8 vacua in Sec. 3, they had no connection
with chiral symmetry -- there was no chiral symmetry for them to
be connected with! Nevertheless, it is an inevitable result of our
analysis, and it explains why all the 8 vacua have the same energy
density and the same expectation value of (F,F); it is because
these quantities are chiral U(l) invariants.
[Parenthetical remark: The factor of 2 in Eq. (5.49) is worth
comment. It tells us that when we make a chiral rotation by TI we
return to the same 8 vacuum. This is as it should be.
-iTIY
e 5 = -1 . (5.50)
Thus a chiral rotation by TI has the same effect on the fields as a

spatial rotation by 2TI; we would be very unhappy if this symmetry
suffered spontaneous breakdown.]
There is one possible loophole in the argument I have given:
It remains a logical possibility that, for every Green's function,
the derivative with respect to a and the derivative with respect to
8 both vanish. If this happened, we would have, not spontaneous
symmetry breakdown, but manifest symmetry, and the 8 vacua would be
mathematical artifacts, superfluous duplicates of a single vacuum.
I will now eliminate this possibility by computing, in the
dilute-gas approximation
f [dA] [d1jJ] [d1)J] e - S e i v8 0+ (x)
(5.51)
f[dA] [d1jJ] [d\i7]e- S e iV8
where
(5.52)
These are chiral eigenfields,
(5.53)
Thus, if we obtain a non-zero answer, we will know that spontaneous

symmetry breakdown has occurred.
The computation will parallel closely that of the vacuum energy
of a pure gauge field theory in Sec. 3. Indeed, as the calculation
proceeds, we will accumulate all the terms that led to our earlier
expression for the determinantal factor, K, as an integral over
instanton size, p,
f,
00
K = 2g- 8 f (pM) ,
o
where M is the renormalization mass. As these old terms come up,
I'll call them to your attention, but I won't bother to write them
down; I'll keep explicit track only of new terms that modify the
integrand in Eq. (5.54).
There is one important novelty in the dilute-gas approximation.
For n widely separated instantons and anti-instantons, i~ has n
vanishing eigenvalues. Thus the integral over Fermi fields will
vanish unless the integrand contains
II a a (5 .55)
r r
A =0
r
Such a term can appear only if we are computing a Green's function
involving at least 2n Dirac fields. Hence, for any fixed Green's
function, the potentially infinite sum over instantons and anti-
instantons terminates.
I will first do the 0 computation:
In the denominator of Eq. (5.51), the only configuration that
does not have a surplus of vanishing eigenvalues is one of no in-
stantons and no anti-instantons, that is to say, the classical
vacuum, Ajl = o. Thus the denominator is simply the product of a
Bose determinant and a Fermi determinant. The same Bose determinant
874 S. COLEMAN
appeared in the denominator in our earlier computation. The Fermi

determinant, det(ia),is a new factor.
In the numerator, we need a configuration with V = I, by Eq.
(5.47). The only one that does not have a surplus of vanishing
eigenvalues is one instanton and no anti-instantons. Let us do the
Fermi integral first; this gives
(5.56)
where det', as always, denotes a determinant with vanishing eigen-

values removed. The Bose integral gives a determinant and a bunch
of collective-coordinate factors identical to those that go into K.
Because det'(i~) does not depend on X, the integration over the
instanton location is trivial,
Jd'+X 1/1; 1/1 0 (x-X) I . (5.57)
_8'IT2/g2
Finally, we have a factor of e from the instanton action,
i8 iv8
and a factor of e from the e
The a+ computation is almost identical to the a one; the only
difference is that the relevant configuration is one anti-instanton,
and thus, instead of a factor of e
is ,we have one of e
-is
Putting all this together, we find
00
<8 ICJ (x) 18> = e -8'IT 2 / g2 e =ti8 -8

g
2J dp f (pM) det ' (i~)
pS det (ia)
o (5.58)
(In case you've lost track of the meaning of my symbols, I remind
you that i~ is the Dirac operator in the field of an instanton of
size p.)
Just as before, we can use dimensional analysis to study the
integrand in this formula. The eigenvalues of i~ have the dimen-
sions of l/length. One eigenvalue has been removed from the primed
determinant; thus the ratio det'/det has dimensions of length, and
must be of the form
det' (iD)
ph(pM) , (5.59)
det (H)
where h is an unknown function. Note that this gives the right

dimensions for the expectation values of o' 1/(length)3.
From here on the argument is a rerun of that of Sec. 3: We can
use the renormalization group to determine the form of the integrand
up to an arbitrary multiplicative constant, be embarrassed in the
infrared, wave our hands about new physics giving an effective in-
frared cutoff, etc.
We now know spontaneous symmetry breakdown occurs. Are there
Goldstone bosons? Let's look for them in
(5.60)
By reasoning which should now be familiar to you, only two field

configurations are relevant: A = 0, and one instanton plus one
)J
anti-instanton. The first of these just gives the usual one-loop
perturbation theory expression; this has a two-quark cut, but no
Goldstone pole. The second just gives the product <8\0+\8><8\0_\8>.
This also has no Goldstone pole. By similar methods one can in-
vestigate other gauge-invariant Green's functions, such as
<8\j5 0 \8> or <8\j5 j 5\8>, and again find no Goldstone poles, but
)J )J'J
really there is no need to do these computations. If Goldstone
bosons appear anywhere, they should appear in (5.60), and they don't.
In the last sentence, I should have said not "appear anywhere",
but "appear among the physical states", that is to say, as singu-
larities in gauge-invariant Green's functions. The situation is
very different if we study a gauge-variant Green's function such as
<e\J~(X)0_(0)\e> = <e\j~(X)0_(0)\e> + l6~2 <e\G w(X)0_(0) Ie> ,

(5.61)
where G is defined in Eq. (3.41). As I have said, the first of
W
the terms on the right has no Goldstone pole, but, as I will show,
the second does. The argument is simple: In a covariant gauge,
there is a Goldstone pole if and only if
fd"X aw<e\Gw(x)0- (0) Ie> of 0 . (5.62)

876 s. COLEMAN
If we use the identity,
(5.63)
and the fact that the only configurations that contribute to (5.62)
have \I = I, we find
!d 4 xd/e IGjl(X)O_(O) Ie> = 32TI 2<elo-,e> of. 0 (5.64)
On the other hand, for <elJ J Ie>, the contributing configura-

jl \I
tions have vanishing \I, and thus there is no Goldstone pole.
To summarize, we have found in the dilute-gas approximation:
spontaneous breakdown of chiral U(l) symmetry, no Goldstone poles
in gauge-invariant Green's functions, no Goldstone poles in the
propagator of a gauge-variant conserved current, and a Goldstone
pole in the Green's function for one gauge-variant current and one
gauge-invariant operator. This is the Goldstone dipole of Kogut
and Susskind.
5.5 QeD (The Real Thingl
Real quantum chromodynamics in the chiral SU(2) ~ SU(2) limit
differs from our baby version in two respects. Firstly, we have
triplet quarks with gauge group SU(3) rather than doublet quarks
with gauge group SU(2). Secondly, we have two massless quarks,
rather than one. [I will ignore the massive quarks; they are
irrelevant to the U(l) problem.]
Replacing an SU(2) doublet by an SU(3) triplet makes hardly
any change. If this were the only difference, we would still have
instantons, and the constant e of Eq. (5.25) would still be ~; the
only thing we would need to change in Sec. 5.4 would be the integral
over instanton size, where g-8 would become g-12 .
In contrast, replacing one massless triplet by two makes a
profound change. e is doubled, and thus the sum rule (5.37) is
changed to
(5.65)
Hence, iD in an instanton field has two vanishing eigenvalues rather

than one. O'l1e don't really need a fancy sum rule to see this; we
have two independent quark fields, so every eigenvalue occurs twice,

once for ~l and once for ~2') Thus, two fields no longer suffice
to take care of all the vanishing eigenvalues, and all quark bi-
linears have zero expectation values.
This is no obstacle to demonstrating the spontaneous breakdown
of chiral U(l) symmetry; we just have to study quadrilinears rather
than bilinears. For example, the same computation that before gave
a non-vanishing expectation value for ~l (1-Y5)~1 will now give a
non-vanishing expectation value for ~1(1-Y5)~1~2(1-Y5)~2'
There is a reason for this. We have found spontaneous break-
down of chiral U(l), but not of chiral SU(2) ~ SU(2); the 8 vacua
are all invariant under chiral SU(2) ~ SU(2). [There are two ways
to see this: (1) There are too few 8 vacua for them to be anything
but invariant; for spontaneous breakdown of chiral SU(2) ~ SU(2)
we need at least a three-parameter family of vacua. (2) Chiral
U(l) transformations are connected to 8 by the anomalous divergence
of the isosinglet axial current; the isotriplet axial current is
anomaly-free.] All Lorentz-invariant quark bilinears transform
according to the representation (~,~) of SU(2) ~ SU(2), and must
have vanishing expectation values. However, there are quadrilinear
SU(2) ~ SU(2) singlets, such as
~E: E: kn 1jJ. (l-y )~k~' (l-y )~n = 1jJ (l-y )~ ~ (l-y)~ -~ (l-y)~ ~ (l-y)~ .
1J IV 1 5 J 5 IV 1 5 1 2 5 2 1 5 2 2 5 1
(5.66)
These operators can have non-vanishing expectation values.
The doubling of C also changes Eq. (5.49) to
(5.67)
Thus a chiral rotation by n/2, rather than n,
~ 1 , 2 -+ -i Y 5 ~ 1 , 2 ' (5.68)
returns us to the same 8 vacuum. Again, this is an effect of un-

broken SU(2) ~ SU(2). If we multiply this by the SU(2) ~ SU(2)
transformation,
878 S. COLEMAN
(5.69)
we obtain
(5.70)
which should not be spontaneously broken.

Of course, we don't want unbroken SU(2) e SU(2) in quantum
chromodynamics; we want spontaneous breakdown; we want pions. How-
ever, there is no reason to be disturbed that pions have not emerged
from our computations. Our methods are semiclassical, valid in the
limit of vanishing~, in principle capable only of revealing those
phenomenon that occur for arbitrarily weak coupling. We have
learned that the breakdown of chira1 SU(2) @ SU(2) is not such a
phenomenon. This is no surprise. What is a surprise (and a wonder-
ful surprise) is that the breakdown of chira1 U(l) is such a
phenomenon.
5.6 Misce11any:J+
There are some topics that I do not have the time to discuss
in the detail they deserve but which I can not resist mentioning:
(1) For most theories with spontaneous symmetry breakdown,
symmetry is restored at sufficiently high temperatures. Is this
true here? This is an easy question to answer. Finite-temperature
Green's functions are given by functional integrals over a Euclidean
time inversely proportional to the temperature, with periodic time
boundary conditions for Bose fields and antiperiodic ones for Fermi
fields. Thus, as the temperature goes up, instantons of any given
size eventually get squeezed out; there's no way to fit them into
the available region of Euclidean space. However, no matter how
high the temperature, there are always instantons so small that they
barely notice the time boundary conditions. Thus, although assym-
metries go to zero as a (calculable) power of the inverse tempera-
ture, symmetry is never fully restored. For extremely high temper-
atures, the only relevant instantons are so small that the effective
coupling constant is extremely weak; thus we could make numerical
computations of extreme accuracy, but only in a regime that is
totally inaccessible to experiment. I stress that this persistence

of symmetry breakdown is a reflection of the scale invariance of
classical chromodynamics, not of any property of instanton effects
in general. For example, in the model of Sec. 4, there is a def-
inite instanton size, and thus, at sufficiently high temperatures,
all instanton effects disappear.
(2) Callan, Dashen, and GrossI have recently proposed a de-
tailed picture of the dynamic structure of quantum chromodynamics.
To explain their ideas, let me restrict myself to chromodynamics
with two massless quarks, and let me imagine the universe cooling
down from a very high temperature. Then, according to Callan,
Dashen, and Gross:
(a) At very high temperatures, when the effective
coupling constant is very small, chiral U(l) is spontaneously
broken by instantons, but chiral SU(2) ~ SU(2) is still a good
symmetry, and quarks are still unconfined. (Of course, this
part is the standard picture which I have described in detail.)
(b) At somewhat lower temperatures, the effective
coupling constant grows larger, and chiral SU(2) ~ SU(2) suf-
fers spontaneous breakdown.~This is also an instanton effect,
but an indirect one that can not be seen in the dilute gas ap-
proximation. Nevertheless, the effective coupling constant,
although not tiny, is still small enough so that weak-coupling
approximations are fairly reliable. (This part looks good to
me.) Quarks are still unconfined.
(c) As still lower temperatures, and still larger ef-
fective couplings, new field configurations, called "merons" ,
become important in the functional integral. These produce a
long-range force that confines the quarks. 36 (I can see nothing
wrong with this idea in principle, but the details of the
argument involve a stupendous amount of hand-waving. This part
is just a suggestion (although a very clever suggestion) that
mayor may not someday become a theory of confinement.)
880 s. COLEMAN
If you will excuse me for beating a dead horse one more time,
this picture shows very sharply how misleading it is to say that
"instantons give the D(l) Goldstone boson a mass". This implies
that quarks get their masses through spontaneous symmetry break-
down, with the appearance of four Goldstone bosons, and then instan-
tons come to the rescue. This is not what happens.
(To be fair, I should modify the last sentence and say, "This
is not what happens in the picture of Callan, Dashen, and Gross."
A skeptic might imagine replacing paragraph (c) above by, "At still
lower tempera~ures, and still larger effective couplings, new field
configurations become important which restore chiral D(l) invari-
ance. At a yet later stage, this suffers spontaneous breakdown and
a Goldstone boson appears." To my knowledge, there is no chromo-
dynamic computation that offers the slightest evidence for this dis-
gusting alternative, but it is not logically excluded.~)
(3) I have stressed several times that spontaneous breakdown
of D(l) (without Goldstone bosons) is independent of spontaneous
breakdown of SD(2) 8 SD(2) (with Goldstone bosons). In a recent
paper, Crewther 38 has argued ingeniously that these phenomena are
not just independent; they are inconsistent. This would be bad
news if it were true, but I do not believe that it is; I think
Crewther's arguments are invalid. However, since Crewther and I
are at this moment entering our fourth month of correspondence on
this matter, and since neither of us has yet convinced the other of
the error of his ways, I will say no more about this.
(4) In all the e vacua, except for e= a or 1f, CP-noninvariant
operators have nonvanishing expectation values. Thus it seems that
in most of the e vacua we have observable strong CP violation. Of
course, this is an illusion; the e vacua are transformed into each
other by the D(l) group, and thus all experiments must yield the
same results in any vacuum. Phrased more explicitly, for every e
vacuum there is a discrete symmetry under which the vacuum is in-
variant, the product of CP and an appropriate D(l) transformation,
THEUSESOFINSTANTONS 881
and we are free to redefine CP to be this transformation.

All this is for massless quarks. The situation changes drastic-
ally when the quarks have masses, either because we have put them
in by hand, or because they have Yukawa couplings to weak-interaction
Higgs mesons. Now we no longer have U(l) symmetry; there is a
potential clash between the definition of CP selected by e and that
selected by the quark mass operator, and there is the disastrous
possibility of strong CP violation.
[Let me dispose of a red herring. You might think that all
this might be said of a theory in which U(l) breaks down in the
ordinary way, with Goldstone bosons, as in the U(l) 0 model. In
this case, there is no problem; as soon as we add a U(l) violating
interaction, no matter how weak, the order parameter, the analog of
e, automatically aligns itself with the perturbation. This is not
what happens here. The easiest (and unfortunately also the least
convincing) way of seeing this is to remember that when all the dust
of Sec. 3 settled, e emerged as effectively a coupling constant, the
coefficient of a term in the action. Thus we would no more expect
e to change discontinuously in response to an external perturbation
than we would expect g to.]
Several mechanisms have been suggested for avoiding this dis as-
ter.~ At the moment I favor an up quark with vanishing bare mass,
that is to say, with vanishing coupling to the Higgs fields. In
this case, we still have a U(l) symmetry, chiral U(l) acting on the
up quark only, and thus we have no CP problem. Unfortunately, this
conflicts with current-algebra estimates of the up mass; these all
agree that it is somewhere between Y2 and ~3 of the down mass. How-
ever, all these estimates are based on soft-kaon and soft-eta com-
putations, and these are notoriously less accurate than soft-pion
computations. For example, only soft-pion methods are needed to
compute the slope of n + 3n, in good agreement with experiment;
soft-eta methods are needed to compute the rate, off by a factor of
three. 4o So perhaps a massless up quark is not such a silly idea.
882 s. COLEMAN
Still, I would be happier if I had a more elegant solution, and one

with more predictive power.
VI. THE FATE OF THE FALSE VACUUH 41
6.1 Unstable Vacua

In Sec. 2.4 I explained how to use instanton methods to study
a particle theory with a false (that is to say, unstable) ground
state. In this section I will apply these methods to a field
theory with a false ground state, that is to say, a false vacuum.
For simplicity, I will restrict myself to the theory of a
single scalar field in four-dimensional space-time, with dynamics
defined by the Euclidean action
S = fd4x[Hd)1<1 2 +U(<I] , (6.1)
where U is a function of the form shown in Fig. 15. Note that U

possesses two relative minima, <1>+ and , but only <1>_ is an absolute
minimum. In analogy to Sec. 2.4, I have used my freedom to add a
constant to U to insure that U(+) = O. The state of the classical
field theory for which = is the unique classical state of lowest
energy, and, at least for weak coupling, corresponds to the unique
vacuum state of the quantum theory. The state of the classical
field theory for which = <1>+ is also a stable classical equilibrium
state. However, in the quantum theory it is rendered unstable by
barrier penetration; it is a false vacuum.
Figure 15
Even without any knowledge of instantons and bounces, it is

easy to understand the qualitative features of the decay of the
false vacuum. The decay closely parallels the nucleation processes
of statistical physics, like the crystallization of a supersaturated
solution or the boiling of a superheated fluid. Imagine Fig. 15
to be a plot of the free energy of a fluid as a function of density.
The false vacuum corresponds to the superheated fluid phase and the
true vacuum to the vapor phase. Thermodynamic fluctuations are con-
tinually causing bubbles of vapor to materialize in the fluid. If
the bubble is too small, the gain in volume energy caused by the
materialization of the bubble is more than compensated for by the
loss in surface energy, and the bubble shrinks to nothing. However,
once in a while a bubble is formed large enough so that it is ener-
getically favorable for the bubble to grow. Once this occurs,
there is no need to worry about fluctuations anymore; the bubble
expands until it converts the available fluid to vapor (or coalesces
with another bubble).
An identical picture describes the decay of the false vacuum,
with quantum fluctuations replacing thermodynamic ones. Once in a
while a bubble of true vacuum will form large enough so that it is
energetically favorable for the bubble to grow. Once this happens,
the bubble spreads throughout the universe, a cancer of space, con-
verting false vacuum to true.
Thus the thing to compute is not a decay probability per unit
time, f, but a decay probability per unit time per unit volume,
f/V, for the probability per unit time that in a given volume a
critical bubble will form is proportional to the volume (at least
i f the volume is much bigger than the bubble) .
Of course, such a computation would be bootless were it not for
cosmology. An infinitely old universe must be in a true vacuum, no
matter how slowly the false vacuum decays. However, the universe
is not infinitely old, and, at the time of the big bang, the uni-
verse might well have been in the false vacuum. For example, in
884 s. COLEMAN
the Weinberg-Salam model, if the mass of the Higgs meson exceeds

Heinberg's lower bound, the asymmetric vacuum, in which we live,
has a lower energy than the symmetric vacuum. However, if the Higgs
mass is less than 1:2 times the lower bound, the symmetric vacuum is
a local minimum of the potential, a possible false vacuum. Now we
know that at high temperatures (i.e., in the early universe),
symmetry breaking disappears in this model; the symmetric vacuum
is the true ground state. Thus it is possible to envision a situ-
ation in which the universe gets into the false vacuum early in its
history and is stuck there as it cools off; in such a situation,
knowledge of f/V is essential if we wish to describe the future
of the universe.
[I stress that I'm just using the Heinberg-Salam model as an
example. I've chosen it because it's familiar and concrete, but in
some ways it's a bad choice for our purposes. Firstly, the model
involves, not one scalar field, but many scalar and vector fields.
Secondly, the vacuum stability features I have described are not
properties of the classical potential, U(), but require considera-
tion of one-loop corrections. Thus the formalism I am going to
develop is not applicable to this case. As long as we're talking
about this model, though, you might be tempted to consider the pos-
sibility that the Higgs mass is less than Heinberg's lower bound,
that we are living in the false vacuum. As Linde 41 has pointed out,
this is silly; if this were the case, there would be no way for the
universe to get into the false vacuum in the first place.]
The relevant parameter for cosmology is that cosmic time for
which the product of f/V and the volume of the past light cone is
of order unity. If this time is on the order of microseconds, the
universe is still hot when the false vacuum decays, even on the
scale of high-energy physics, and a zero-temperature computation of
f/V is inapplicable. If this time is on the order of years, the
decay of the false vacuum will lead to a sort of secondary big bang,
with interesting cosmological consequences. If this time is on the
order of billions of years, we have occasion for anxiety.

6.2 The Bounce
We know from Sec. 2.4 how to compute r/v. We must find the
bounce, ~, a solution of the Euclidean equations of motion,
(6.2)
that goes from the false ground state at time minus infinity to the
false ground state at time plus infinity,
To these boundary conditions we can add another. It is easy to see

that if the action of the bounce is to be finite,
- -+
lim ~(x,x ) = ~+ . (6.4)
I~I-+oo It
Once we have found the bounce, it is trivial to compute r/v. To

leading order in -ti,
(6.5)
where So is S(~) and K is a determinanta1 factor, defined as in Sec.

2.4.
I will shortly construct the bounce. Before I do so, though,
I want to make some comments:
(1) We already see the power of our method. The problem of
barrier penetration in a system with an infinite number of degrees
of freedom has been reduced to a study of the properties of a
single classical partial differential equation.
(2) The factor of V in the expression for r arises automatic-
ally in our method. No non-trivial solution of Eqs. (6.2)-(6.4) is
translation invariant. Thus we must integrate over the location of
the bounce. This gives us a factor of V, just as did the integra-
tion over instanton location in Sec. 3.
(3) It might be that there are many solutions to Eqs. (6.2)-
(6.4). We are only interested in the solutions of minimum action,
for these make the dominant contribution to the functional integral.
886 s. COLEMAN
We are not interested in the trivial solution, = +.
(4)
For this solution, 8 2S/82 has no negative eigenvalues, and thus
makes no contribution to the vacuum decay probability.
(5) If we imbed in a one-parameter family of functions,
(6.6)
then,
(6.7)
Because is a solution of the equations of motion, this must be

stationary at A = 1. Thus,
fd 4X(d].l{j))2 = - 4fd 4XU({j)) , (6.8)

and
(6.9)
This is reassuring. Since U is somewhere negative, one might worry

about the possibility that So was negative, which would lead to a
very strange dependence of the decay probability on~. This pos-
sibility has now been eliminated. Also,
d 2S/dA 2 = -tJd4X(d].l{j)) 2 < 0 . (6.10)
Thus, at (j), 8 2S/82 has at least one negative eigenvalue, and (j) does
contribute to the decay probability. Of course, if there were more
than one negative eigenvalue, we would have to rethink the analysis
of Sec. 2.4. However, as I shall show eventually, this does not
happen; there is only one negative eigenvalue.
Now for the construction of the bounce: Eqs. (6.2)-(6.4) are
0(4) invariant. Thus it is not unreasonable to guess that the
bounce might also be 0(4) invariant, that is to say, that might
depend only on the distance from some point in Euclidean space.
Recently, Glaser, Martin, and I were able to show that this guess
is right, under mild conditions on U; there always exists an 0(4)-
invariant bounce and it always has strictly lower action than any
0(4)-noninvariant bounce. 42 The rigor of our proof is matched only
by its tedium; I wouldn't lecture on it to my worst enemy. However,

it is possible to give a sloppy argument for the first part (exist-
ence) although, unfortunately, not for the second (action minimiza-
tion).
I will now give this argument.
If we choose the center of symmetry to be the origin of co-
ordinates, then 0(4) symmetry is the statement that is a function
only of the radial variable, r. Thus Eq. (6.2) becomes
d 2 +1 Ei = u' () (6.11)
dr 2 r dr '
while Eqs. (6.3) and (6.:4) both become
lim
r-+oo
(r) = + (6.12)
Also,
(6.13)
Otherwise, would be singular at the origin.

The key to the argument is the observation that if we interpret
as a particle position and r as time, Eq. (3.9) is the mechanical
equation for a particle moving in a potential minus U and subject to
a somewhat peculiar viscous damping force with Stoke's law coeffi-
cient inversely proportional to the time. The particle is released
at rest at time zero, Eq. (6.13); we wish to show that if the ini-
tial position is properly chosen, the particle will corne to rest at
time infinity at +, that is to say, on top of the right-hand hill
in Fig. 16.
-u
Figure 16
888 s. COLEMAN
I shall demonstrate this by showing that if the particle is re-

leased to the right of ~_, and is sufficiently close to ~_, it will
overshoot and pass ~+ at some finite time. On the other hand, if it
is released sufficiently far to the right of ~ , it will undershoot
and never reach ~+. Thus (arguing in the worst tradition of nine-
teenth century British mathematics) by continuity there must be an
intermediate initial position for which it just comes to rest at ~+.
To demonstrate undershoot is trivial. If the particle is re-

leased to the right of ~o' it does not have enough energy to climb
the hill to ~+. The damping force does not affect this argument,
because viscous damping always diminishes the energy.
To demonstrate overshoot requires a little more work. For ~
very close to ~_, we may safely linearize Eq. (6.11),

d2 3 2-
(dr 2 + ~ - ].1 ) (~ - ~_) = 0 , (6.14)
where ].12 is U" (~ ). The solution to Eq. (6.14) is
(6.15)
Thus, if we choose ~ to be initially sufficiently close to ~_, we

can arrange for it to stay arbitrarily close to ~ for arbitrarily
large r. But for sufficiently large r, the viscous damping force
can be neglected, since it is inversely proportional to r. But if
we neglect viscous damping, the particle overshoots. Q.E.D.
We have made great progress. We have reduced the partial dif-
ferential equation for the bounce to an ordinary differential equa-
tion. But we can go even farther; in the limit of small energy-den-
sity difference between the true and false vacuum, we can obtain an
explicit expression for the bounce and for So' as I shall now show.
6.3 The Thin-Hall Approximation
Let U+(~) be an even function of ~,
u+(~) = u+(-~) , (6.16)
with minima at some points fa,
U~(a) = 0 . (6.17)
Also, let us define
].12 = U"(a) (6.18)
+
Now let us add to U+ a small term that breaks the symmetry,
U = U+ + E( - a)/2a , (6.19)
where E is a positive number. This defines a theory of the sort we

have been discussing. To lowest non-trivial order in E,
(6.20)
and E is the energy-density difference between the true and the

false vacuums.
It is easy to see the qualitative form of the bounce in the
limit of small E from the mechanical analogy of Sec. 6.2. In order
not to lose too much energy, we must choose (O), the initial
position of the particle, very close to . The particle then stays
close to until some very large time, r = R. Near time R, the
particle moves quickly through the valley in Fig. 16, and slowly
comes to rest at + at time infinity. Translating from the mechan-
ical analogy back into field theory, the bounce looks like a large
four-dimensional spherical bubble of radius R, with a thin wall
separating the false vacuum without from the true vacuum within.
To go on, we need more information about the wall of the bubble.
For r near R, we can neglect the viscous damping term and we can
also neglect the E-dependent term in U. We thus obtain
(6.21)
This is the classical equation of motion for a particle in a sym-

metric double-welled potential, the equation we studied in Sec. 2.2,
the equation that had one-dimensional instantons for its solutions.
Indeed, a one-dimensional instanton centered at R is the solution
we need here, for such a function goes from -a to a as r increases
through R, just what we want. This is our approximate description
of the bounce.
The only thing missing from this description is the value of R.
This is easily obtained by a variational computation:
00
S = 2TI 2 J r3dr[~(d/dr)2 + U] . (6.22)

o
890 S. COLEMAN
We can divide this integral into three regions: the outside of the
bubble, the skin of the bubble, and the inside of the bubble.
Within the accuracy of our approximation, in the outside region,
~ = ~+ and U = 0; thus we get no contribution from this part of
the integral. In the inside region, ~ = ~ and U = -E; thus from
this part of the integral we get
(6.23)
Over the skin, r is approximately R, and, over this small region,

the E-dependent terms in U are negligible; thus from this part of
the integral we get
21T2 R3 Idr[Hd(j)/dr) 2 + U+] = (6.24)
where S1 is the action of a one-dimensional instanton,
-a
S1 = I
a
Izu+ d~ (6.25)
(6.26)
Varying with respect to R, we find
(6.27)
Hence,
(6.28)
This completes the approximate description of the bounce. We also

know so: (6.29)
I have described what we have done as an approximation that is

valid in the limit of small E. Now that we have gone through the
computation, we can phrase the condition for the validity of the
approximation more precisely: the approximation is good if the
radius of the bubble is much larger than the thickness of the bubble
wall; R must be much larger than l/~, or, equivalently,
(6.30)
6.4 The Fate of the False Vacuum

In a particle problem like that of Sec. 2.4, we can describe
the decay process in the language of ;he old quantum theory. The
particle sits at the bottom of the potential well until, at some
random time, it makes a quantum jump to the other side of the bar-
rier, materializing at the point labeled a in Fig. 7. At this
point, the potential energy of the particle is the same as it was
at the bottom of the well; thus its kinetic energy must vanish;
equivalently, it has zero velocity. These conditions give the
initial-value data for the subsequent motion of the particle, which
is totally governed by classical mechanics. Like all descriptions
of quantum-mechanical processes in the language of the old quantum
theory, this one must be taken with a large grain of salt; it will
certainly lead us astray if we try to use it to describe meaure-
ments made just outside the potential barrier. Nevertheless, it is
very useful as an asymptotic description, for discussing what hap-
pens far from the barrier and long after the time the system decays.
For example, this is the description we all use when we discuss the
macroscopic detection of an alpha particle emitted by an unstable
nucleus.
This description can readily be extended to a system with many
degrees of freedom. The point a becomes the point in multi-dimen-
sional configuration space where all velocities vanish; that is to
say, it is the midpoint of the bounce. Thus, for the field theory
we have been studying, the description of the vacuum decay process
in the language of the old quantum theory is: The classical field
makes a quantum jump (say at time zero) to the state defined by
(6.31a)
and
(6.31b)
Afterwards, it evolves according to the classical Minkowskian field

equation,
(6.32)
892 S. COLEMAN
The first of these equations implies that the same function,

(r), that gives the shape of the bounce in four-dimensional Euclid-
ean space also gives the shape of the bubble at the moment of its
materialization in ordinary three-space. Indeed, it does more;
because the Minkowskian field equation is simply the analytic con-
tinuation of the Euclidean field equation back to real time, the
desired solution of Eqs. (6.31) and (6.32) is simply the analytic
continuation of the bounce:
-+ -
(x o ,x) = (r = vlx
J'-+12-x 2 ) (6.33)
o
[As a consequence of Eq. (6.13), is an even function of r, so we

need not worry about which branch of the square root to take.]
We can immediately draw some very interesting consequences of
Eq. (6.33):
(1) 0(4) invariance of the bounce becomes 0(3,1) invariance
of the solution of the classical field equations. In other words,
the growth of the bubble, after its materialization, looks the same
to any Lorentz observer.
(2) In the case of small E, discussed in Sec. 6.3, there is a
thin wall, localized at r=R, separating true vacuum from false.
As the bubble expands, this wall traces out the hyperboloid
l-+xl2 _ Xo2 = R2 (6.34 )
Typically, we would expect R to be a microphysical number, on the

order of a fermi, give or take ten orders of magnitude. This means
that by macrophysical standards, once the bubble materializes it
begins to expand almost instantly with almost the velocity of light.
(3) As a consequence of this rapid expansion, if a bubble were
expanding toward us at this moment, we would have essentially no
warning of its approach until its arrival. This is shown graphic-
ally in Fig. 17. The heavy curve is the bubble wall, Eq. (6.34).
A stationary observer, 0, cannot tell a bubble has formed until he
intercepts the future light cone, W, projected from the wall at the
time of its formation. A time R later, that is to say, on the order
Figure 17
of 10- 10 _10- 30 sec. later, he is inside the bubble and dead. (In the
true vacuum, the constants of nature, the masses and couplings of
the elementary particles, are all different from what they were in
the false vacuum, and thus the observer is no longer capable of
functioning biologically, or even chemically.) Since even 10- 10 sec.
is considerably less than the response time of a single neuron,
there is literally nothing to worry about; if a bubble is coming
toward us, we'll never know what hit us.
(4) The rapidly expanding bubble wall obviously carries a lot
of energy. How much? A section of bubble wall at rest carries
energy Sl per unit area. Because any part of the bubble wall at any
time is obtained from any other part by a Lorentz transformation, a
section of wall expanding with velocity v carries energy Sll/l -v 2
per unit area. Thus, at a time when the radius of the bubble is
I~I, the energy of the wall is
(6.35)
894 s. COLEMAN
By Eq. (6. 34) ,
v (6.36 )
Thus,
(6.37)
Thus, in the thin-wall approximation, all the energy released by

converting false vacuum to true goes to accelerate the bubble wall.
This refutes the naive expectation that the decay of the false
vacuum would leave behind it a roiling sea of mesons. In fact, the
expansion of the bubble leaves behind only the true vacuum.
6.5 Determinants and Renormalization
I said earlier that the determinantal factor K in Eq. (6.5)
was defined as in the particle problem of Sec. 2.4. This is basic-
ally true, but there are three technical differences: (1) In par-
ticle mechanics, we had only one infinitesimal translation, and
thus one zero eigenvalue, to worry about; here we have four.
(2) It was critical in the analysis of Sec. 2.4 that the second
variational derivative of the action at the bounce had one and only
one negative eigenvalue. Is the same true here? (3) Whenever we
study a relativistic field theory, we must deal with ultraviolet
divergences and renormalization. Of course, this last remark also
applied to the gauge field theories of Sec. 3, where I swept re-
normalization problems under the rug. However, we now have a prob-
lem with a much simpler renormalization structure (only a single
scalar field to worry about, no problems with gauge invariance and
gauge-fixing terms, etc.), so it's worth confronting renormalization
head-on.
I will deal with these three problems in the order in which I
have stated them.
(1) Vanishing Eigenvalues. Because we have four infinitesimal
translations, we have four eigenfunctions with eigenvalue zero,
proportional to a~~. We must determine the constant of proportion-
ality, that is to say, the normalization of the eigenfunctions.
This is easy to do. By the spherical symmetry of the bounce,
fd 4 Xd (jid (ji
]l V
(6.38)
by Eq. (6.9).
Thus, as far as zero eigenvalues go, the only difference be-
tween the problem at hand and the particle problem of Sec. 2.4 is
!.:
that we have four factors of (So/2n) 2 rather than one. Hence,
det' [-d]ld]l + U" (ji)] 1-

1
I det
7z
[-d]ld]l+U"(<jJ~] (6.39)
assuming we have no problems with negative eigenvalues or renormal-

ization.
(2) Negative Eigenvalues. We already know that 02S/0<jJ2 evalu-
ated at the bounce has at least one negative eigenvalue. Can there
be more than one? To answer this question I will have to steal
some information from the paper by Glaser, Martin, and I that I
referred to earlier. 42 There we showed that the bounce could be
characterized as the absolute minimum of S for fixed
V= fd 4 XU (6.40)
This implies that there can not be two independent eigenvectors

with negative eigenvalues; for, if there were, we could form a
linear combination of the eigenvectors tangent to the surface of
constant V, and the bounce would not even be a local minimum of S
with fixed V, let alone an absolute minimum.
(3) Renormalization. Until now all of our dynamics has been
expressed in terms of unrenormalized quantities. We must now re-
cast our formulas in terms of renormalized quantities. We begin
with S itself, ~ (n)
S = SR + L S (6.41)
n=l
Here SR is the renormalized action, a functional of exactly the

same form as S, but with all unrenormalized quantities replaced by
their renormalized counterparts, and S(n) is the action induced by
standard renormalization counterterms computed from the sum of all
896 S. COLEMAN
n-loop graphs. To avoid excessive clutter in my equations, I will

redefine to be the renormalized field, U to be the polynomial
that occurs in SR' ~ to be the bounce as computed from SR' and So
to be SR (~) .
The renormalization counterterms serve to remove all ultra-
violet divergences from all one-particle irreducible Green's func-
tions. Equivalently, they serve to remove all ultraviolet diver-
gences from the effective action, Y(), the generating functional
of these Green's functions. To one-loop order,43
k
x det[-Cl Cl +U"()]2 (6.42)
].l ].l
It will be important to us shortly that (for renormalizable U's)

the right-hand side of this equation is free of ultraviolet diver-
gences for arbitrary .
Now let us imagine computing f/V iteratively, first treating
SR as if it were the total action, and then taking account of the
renormalization counterterms perturbatively. If we had not set~
(n) . n
equal to one, S would have been proport10nal to~. Thus, to
the order in which we are working, the only counterterm we need
consider is S(l) .
The first thing we must realize is that the counterterms may
destroy our convention that S(+) vanishes. We can take care of
this trivially by replacing So in Eq. (6.5) by the difference
So - S(+).
Secondly, adding new terms to SR will chahge the stationary
points of S. In particular, it will change the bounce. Let us
write
(6.43)
Then
+ .. , , (6.44)
where the triple dots indicate terms that are negligible in the
order in which we are working. The second term vanishes because
the bounce is a stationary point of SR. Thus, for our purposes,
S(~) So + SCl)C) (6.45)

By the same reasoning,
(6.46)
f/V = 4:
S2
2 exp[-So - S(I) () +SCl) (~+)]
det' [-Cl Cl + U"()] I-~

11 11
det [-Cl Cl + U"(~+)] (6.47)
11 11
The point of this exercise is not the simplicity of this formula.

Equation (6.47) is an ugly mess, and I know no way of evaluating
it for even the simplest theories without using a computer. Rather,
the point is that ordinary renormalization works for instanton
computations: As a good renormalized expression should be, Eq.
(6.47) is free of ultraviolet divergences; each determinant is
paired with an exponential of S(l), just as in Eq. (6.42). (That
one of the factors is a primed determinant is irrelevant; omitting
any finite number of eigenvalues has no effect on the ultraviolet
divergence. )
6.6 Unanswered Questions
This concludes what I know about the fate of the false vacuum.
There remain many interesting unanswered questions:
(1) I have discussed the expansion of a bubble of true vacuum
into false vacuum. What if the initial state of the world is not
the false vacuum, but some state of nonzero particle density built
on the false vacuum? vfuat happens when a bubble wall encounters a
particle?
(2) I have discussed spontaneous decay of the false vacuum.
However, there is also the possibility of induced decay. In partic-
ular, in a collision of two particles of very high energy, there
might be a non-negligible cross section for the production of a
bubble. How can one estimate this cross section?
898 S. COLEMAN
(3) If we assume that the universe starts out in a false

vacuum, at some time in its expansion bubbles begin to form. Be-
cause the formation of bubbles is totally Lorentz invariant, the
average distance between bubbles at their time of formation must
be of the same order of magnitude as the time at which bubbles
begin to appear. Because bubble walls expand with the speed of
light, after a time interval of the same order of magnitude, bubble
walls begin to collide. What happens then? Can such events be
accommodated in the history of the early universe?
The preceding paragraphs are taken verbatim from a paper I
wrote at the end of 1976. I still don't know the answers to any
of these questions; maybe you'll be able to do better than I.
Appendix A
HOH TO COMPUTE DETERMINANTS""
He wish to study the equation
(-a~ +H)lj! = Alj! , (A.l)
where H is some bounded function of t. Let us define lj!A (t) as the

solution of this equation obeying the boundary conditions
lj!A (-T/2) =0 , (A.2)
The operator _a 2 +H (acting on the space of functions vanishing at

t
T/2) has an eigenvalue, An' if and only if
lj!A (T/2) = 0 . (A.3)

n
As in the text, we define
det(-a 2 +H) = IT A (A .4)
t n n
(1) (2) ,,,,(1,2)

Now, let Hand H be two functions of t, and let o/A
be the associated solutions of Eq. (A.l). I will prove that
det [-a~ +H(lt Al = lj!~1) (T/2) .

-a~ +\V_(2t \l lj!~2) (T/2) (A.5)
Proof: The left-hand side of this formula is a meromorphic function

--.. (1) (2)
of A, wlth a slmple zero at each A and a simple pole at each A .
n n
By elementary Fredholm theory, it goes to one as A goes to infinity
in any direction except along the positive real axis. The right-
hand side is a meromorphic function with exactly the same zeroes
and poles. By elementary differential-equation theory, it also goes
to one in the same limit. Thus the ratio of the two sides is an ana-
lytic function of A that goes to one as A goes to infinity in any
direction except along the positive real axis. That is to say, it
is one. Q.E.D.
If we define a quantity N by
900 S. COLEMAN
det(-a 2 +H)
t 2
-\jJ""'o-'(=T'::'/""'2):-- = 'IT 11 N (A.6)
then, by Eq. (A.S), N is independent of W. I will use this expres-
sion to define the normalization constant N in the functional in-

tegral. (Note that no explicit definition of this quantity was
given in the text, so I am perfectly free to define it as I wish
here.) Thus we have the desired formula for evaluating Gaussian
functional integrals,
(A.7)
As a specific example, for the harmonic oscillator, H
\jJo = w- 1 sinhw(t +T/2) , (A.S)
from which Eq. (2.16) immediately follows.

Appendix B
THE DOUBLE HELL DONE DOUBLY HELL"5
In this appendix I shall show that the formulas derived in the

text for the splitting of the ground-state energies in a double-well
potential, Eqs. (2.31) and (2.41), are equivalent to the results of
ordinary wave mechanics. To do this, I will have to both evaluate
the determinants that appear in Eq. (2.41) (using the method of Ap-
pendix A) and do the wave-mechanical computation. To keep my equa-
tions as simple as possible, I will choose my units such that w = 1.
Evaluating Determinants
We have to evaluate a primed determinant, one with the zero
eigenvalue omitted. I will do this by evaluating the full deter-
minant on a finite interval, [-T/2, T/2], dividing this by its
smallest eigenvalue, Ao ' and then letting T go to infinity.
Thus we must construct solutions of
(B .1)
We already know one solution with A = 0,
+ Ae -I t 1 ,
The constant A is determined by the integral expression for the in-
stanton, Ea. (2.21),
t = rdX(2V)-~
= -~n[S:~A-I(a-~)]
+ O(a-x) (B.3)
o
Eauation (B .1) must have a second solution with A = 0, which I
denote by YI' It will be convenient to normalize YI such that its
l.Jronskian with Xl is given by
XldY-YdX (B .4)
t I I t I
Thus,
(B.5)
902 S. COLEMAN
We can now construct ~o of Appendix A. For large T,

-1 T/2 -T/2
~o(t) = (2A) (e xI +e YI) (B.6)
Hence,
~o(T/2) = 1 . (B.7)
This takes care of the determinant. To find the lowest eigen-

value, we must find ~A (t) for small A. This can be done by a
standard method: we turn Eq. (B.l) into an integral equation and
iterate once. This can readily be seen to yield
t
~ 0 (t) - A(2A 2) - I f d t ' [y I (t) X I (t ' ) - X 1 (t) YI (t ' ) ] ~ 0 (t ') ,
-T/2 (B.8)
plus terms of order A2 , which we neglect. By Eq. (B.6),
T/2
~A(T/2) = l-A(4A 2)-lf dt[eTx~-e-Ty~]. (B.9)
-T/2
For large T, the second term in this expression is bounded, and
thus negligible compared to the first term. Thus, for large T,
(B.10)
because xI is properly normalized.

Thus the lowest eigenvalue is given by
(B.ll)
and, for large T,

det' [ - d 2 + u" (~)
t
1 ~o(T/2) 1
Ao e T/2 = 2A 2 (B .12)
Reassuringly, this is non-zero and T-independent.

Plugging this in to Eqs. (2.31) and (2.41), we find that the
lowest energy levels are given by
(B.13)
Solving the Schrodinger Equation
We wish to study the solutions of
- t-ti. 2 ax2 1jJ + v 1jJ = E 1jJ (B.14)
As long as x is not near the bottoms of the wells, we can use stan-
dard yOCB solutions. Near the bottom of each well, though, there
are two turning points. These are not separated by many wavelengths,
so we can not use the standard connection formulas for a linear
turning point. Fortunately, near the bottom of a well, in a region
that includes both turning points, we may safely approximate V by a
harmonic-oscillator potential. Thus, for example, for x near a,
we may write
(B .15)
Our strategy will be to match WKB solutions of Eq. (B.14) outside

the walls to solutions of Eq. (B.lS) in the bottoms of the wells.
Furthermore, since we know the solutions are either even or odd,
we can restrict ourselves to positive x, and only have to do this
awkward matching for the right-hand well.
I will begin by constructing the even and odd lOCB solutions
for 0 ~x < a. If we define
k(x) = [2(V-E)]~ , (B.16)
then these are
k -~ [exp-tr- 1JXk dx' exp _11:- 1JXk dx' ] (B .17)
o 0
For the solutions we are interested in, E is itself of order 11:.

Thus we may ignore E in the fact~r of k-~, and expand to first
order in the exponential,
k = (2V)~ - E(2V)-~ . (B.18)
To match on to the solutions of Eq. (B.lS), we need the form

of the WKB solutions as x enters the regime of validity of the
quadratic approximation to V, V= (a-x)2/ 2 In this regime k(x) is
just (a-x), while we can compute the E-independent term in the in-
tegral by
904 S. COLEMAN
Ix
o
dx(2V)'2
1
fadx(2V)~1 -
o
JXdx(2V)~
a
= t So - Ha-x)2 (B.19)
For the E-dependent term in the integral, we can use Eq. (B.3).
Thus we obtain
1JJ (a-x) -~{ exp..tC 1 [t So - Ha-x) 2
+E R-n So-k2 A- 1 (a-x)]
exp _-tr- 1 [t S - Ha-x) 2 + E R-n S-~ A-I (a-x)]}

o 0 (B.20)
If we write
E =-tt(t+) , (B.21)
then Eq. (B.20) becomes
1JJ { e S 0 /2.lfi So-~ A-~ exp[-(a-x) 2 /2-l'f]
(a-x) -1 e -So /Ut S~ A~ exp [(a-x) 2 /2J6.]}
x [l+O()] (B.22)
We will hold this expression in reserve while we go on to study the

solutions of Eq. (B.15).
We already know one solution of Eq. (B.15), for 0,
1JJ 1 = exp[-(a-x)2/2~] . (B .23)
Of course, there is another (odd, increasing) solution, ~1' This

does not have a simple form in terms of elementary functions, but
its asymptotic form, for Ix-al ~, is easily computed by the WKB
approximation, or just read off from Eq. (B.22),
It will turn out that this is all that we need. Note that I have
normalized ~1 such that the Wronskian of the two solutions is
(B.25)
He wish to solve Eq. (B.15) for small E. By the same arguments

as led to Eq. (B.B),
f dx' \jJ
00
\jJ = \jJ 1 1 (x' ) [\jJ 1 (x') <P 1 (x) - <P 1 (x') \jJ 1 (x) ]
- E (B.26)
x
I have chosen here the solution that vanishes as x goes to plus in-
finity. Thus, this is the appropriate solution for matching with
the decreasing WKB solution in the region (x-a) -ti. Thus, the
only matching left to do is in the region (a-x) ~.
In this region, we can use

00
f dx \jJ~
'-00
;.rr::ti (B.27)
to write
\jJ exp [ - (a-x) 2 /2tl:] [HO (E) ] - E ('Tf11:) \a-x) -1 exp [ (a-x) 2/ 2tr ] (B .28)
As it should be, this is proportional to Eq. (B.22), if we choose

e -s 0 /-tr A(S /1H'r) ~ (B.29)
o
This is the desired result, and it is identical to the result of
the dilute-gas approximation, Eq. (B.13).
Almost identical methods to these can be used to check the
dilute-gas formula for the width of an unstable state, Eq. (2.50).
Yo~ might find it an instructive exercise to see that things work
out in this case also.
906 s. COLEMAN
Appendix C
FINITE ACTION IS ZERO MEASURE 46
In this appendix I will show that, even for a one-dimensional

harmonic oscillator, motions of finite action form a set of measure
zero in function space.
If we define eigenvalues An and expansion coefficients c
n
as
in Sec. 2.1, then, for a harmonic oscillator, the quadratic approxi-
mation to the action is exact,
S = -!- \" A c 2 (C.l)
L n n
n
If we introduce new variables, b

n
= c
n
IA n /~, then
S =-tr/2 Lb n2
n
(C.2)
Let us define a slightly unconventional normalization constant, N',

by
(C.3)
This has been chosen such that
N'f[dXle-S/~ = 1 (C.4)
How much of this integral comes from motions of finite action?

The integrand is positive, and every motion of finite action lies
in a cube of side L
for all n , (C.S)
for sufficiently large L. Thus, the finite-action contribution to

the integral must be less than
lim ; (2n)
L-><Xl n=l
-~f~ b n e -b~ o . (C.6)
-L
Q.E.D.
Appendix D
ONLY WINDING NUMBER SURVIVES
This appendix is the promised (in Sec. 3.3) demonstration that,

for a sufficiently large box, the only relic of the boundary condi-
tions imposed on the walls of the box is the winding number.
Consider a rectangular box in Euclidean four-space, with sides
L 1 L 4 I will label the eight hyperplanes that bound the box by
their normal vectors; thus I will refer to the upper l-wall, the
lower l-wall, the upper 2-wall, etc. (Upper and lower here refer
to greater and lesser values of the appropriate coordinate.)
On the walls of the box the tangential components of A are
]J
given in a way consistent with finiteness of the action, that is to
say, consistent with A = g d g -1 (D .1)
]J ]J
Thus, giving the tangential components of A on the walls is equiv-

]J
alent to giving g on the walls (up to an irrelevant multiplicative
constant). The gauge condition A3 = a still allows arbitrary x 3-
independent gauge transformations. I will use the freedom to make
such a transformation to transform g to one on the lower 3-wall.
Because the vanishing of A3 implies the vanishing of d 3 g, g is
automatically one on all walls except the upper 3-wall. On this
wall, g is given as a function of three variables, g(x 1 , x 2 , x 4 ) ,
equal to one on the boundary of the wall. (I stress that the only
function of this gauge transformation is to simplify my subsequent
arguments. Since the functional integral is gauge-invariant, any-
thing I can prove with this gauge convention I could prove without
it; it's just that the arguments would be clumsier.)
Now let us imbed our original box, with boundary conditions
given by gl(x 1 , x 2 , x 4 ) in a larger box, with the same lowermost
corner (chosen to be the origin of coordinates), and with the same
sides Ll' L 2 , and L4 , but with third side L3+L'I. Let the boundary
conditions on the larger box be given by some function g2(x 1 , x 2 , x4 ).
908 S. COLEMAN
Theorem: If gl and g2 are in the same homotopy class, then

any field configuration defined inside the original box consistent
with its boundary conditions can be extended to a field configura-
tion defined inside the larger box, consistent with its boundary
conditions and the gauge condition A3 =0, at the cost of an increase
in action of order Ij~.
Before I prove this theorem I will make some comments:
(1) The theorem would certainly not be true if gl and g2 were
in different homotopy classes. In this case, to get from gl to g2'
we would have to put at least one instanton in the new volume; this
would increase the action by at least 8n 2 jg2, independent of the
value of ~.
h
(2) We are free to choose ~ to be proportional to, say, L;.
Thus, for a very large box, the fractional change in the volume of
the box is negligible, as is the change in the action. In the
language of statistical physics, changing the boundary conditions
while keeping the winding number fixed is just a surface effect, not
a volume effect.
(3) There is an apparent paradox that may have bothered you:
For any fixed configuration of instantons and anti-instantons,
g(x 1 , x 2 ' x 4 ) is fixed. How then can we get all configurations
consistent with a fixed winding number with a single set of bound-
ary conditions? The theorem supplies the answer: We don't get all
these configurations; we get only a small portion of them. However,
we do get "close relatives" of all of them, configurations that dif-
fer only by a small distortion very close to the upper 3-wall. The
difference caused by this small distortion is negligible for a suf-
ficiently large box.
Now for the proof: By assumption, g and g' are in the same
homotopy class. Thus there is a continuous function of four vari-
ables, g(xl' x 2 ' s, x 4 ) , with 0 Ss sl, such that
g(x , x , 0, x ) = g , g(x , x , 1, x ) g . (D.2)

1 2 4 1 1 2 4 2
Let g(x) be a function defined in the added volume by
(D.3)
If we could choose
A = g d g-1 (D.4)
)J )J
then we could effect the desired transition at no cost in added

action. Unfortunately, this is impossible; Eq. (D.4) is inconsist-
ent with the gauge condition A3=O. However,
A d -1
)J 3
)J g )J g ofo
= 0 , )J = 3 , (D.S)
is consistent with the gauge condition and will effect the transi-
tion.
We must compute the action associated with Eq. (D.S). If we
make a gauge transformation by g-l, Eq. (D.S) becomes
A
)J
o )J ofo 3
g-1 d)J g , )J = 3 . (D.6)
(A gauge transformation does not change the action.) From Eq.

(D.6), we see that A3 is proportional to 1/6. The only non-vanish-
ing components of F)Jva r e F)J3 , also oroportional
,- to 1/6. Thus the
Lagrangian density is proportional to 1/6 However, the volume of
2
integration is only proportional to 6. Q.E.D.

910 s. COLEMAN
Appendix E
NO HRONG-CHIRALITY SOLUTIONS 47
In this appendix I will show that, if
F =
-
F (E .1)
~\! ~\!
then the only normalizable solution of both
o (E.2)
and
\jJ , (E.3)
is \jJ = O.
From Eq. (E .2) ,
o. (E.4)
Also,
F~\! Y].l Y\! Y 5 (E.5)

Thus,
D D \jJ = 0 (E .6)
~ ].l
Hultiplying by \jJ+ and integrating, we find
Jd 4 XD \jJ+D \jJ
~ ~
= 0 (E.7)
Hence
D ,I, = 0 (E.8)
~ 't' ,
for all~. If we go to axial gauge, this implies, in particular,

that \jJ is independent of x 3 The only such normalizable function
is \jJ = O. Q.E.D.
NOTES
1. These topics are all drawn from the classic part of the theory.
"Classic",in this context, means work done more than six months
ago. A good summary of the more recent research of one of the
most active groups in this field is C. Callan, R. Dashen, and
D. Gross, "Toward a Theory of the Strong Interactions" (to be
published in Phys. Rev.).
2. Polyakov's early work is summarized in A. M. Polyakov, Nucl.
Phys. B 121, 429 (1977).
3. See, for example, R. Feynman and A. Hibbs, Quantum Mechanics
and Path Integrals (McGraw-Hill, New York, 1965).
4. See the note on notation at the end of Sec. 1.
5. It was Polyakov (Ref. 2) who recognized the double well as the
prototypical instanton problem.
6. For a review of lumps, see my 1975 Erice lectures, "Classical
Lumps and Their Quantum Descendants", in New phenomena in Sub-
nuclear Physics, ed. by A. Zichichi (Plenum Press, New York,
1977) .
7. This is, of course, nothing but the standard prescription for
handling collective coordinates in soliton problems. See
J. L. Gervais and B. Sakita, Phys. Rev. D1l.., 2943 (1975).
8. The treatment here follows that of C. Callan and S. Coleman,
Phys. Rev. Dli, 1762 (1977). The idea of handling unstable
states this way goes back to Langer's analysis of the droplet
model in statistical mechanics [J. S. Langer, Ann. Phys.(N.Y.)
~, 108 (1967)]. The factor of ~, of which much is made below,
occurs in Langer's analysis and was explained to me by Michael
Peskin.
9. The order of my exposition will not be the historical order of
discovery. Here is the way it happened: The topological struc-
ture of finite-action Euclidean gauge-field configurations was
uncovered and the instanton solutions discovered by A. A.
Belavin, A. ti. Polyakov, A. S. Schwartz, and Yu. S. Tyupkin,
912 S. COLEMAN
Phys. Lett. 59B, 85 (1975). The imoortance of the instantons

was realized by G. 't Hooft [Phys. Rev. Lett. 12, 8 (1976);
Phys. Rev. D~, 3432 (1976)] who used them to solve the U(l)
problem. (I won't get to this until Sec. 5.) 't Hooft's work
was clarified and extended by R. Jackiw and C. Rebbi [Phys. Rev.
Lett. 12, 172 (1976)] and by C. Callan, R. Dashen, and D. Gross
[Phys. Lett. 63B, 334 (1976)], who discovered the properties of
pure gauge field theories discussed in this section.
10. For a review of gauge field theories, see my 1973 Erice lec-
tures, "Secret Symmetries", in LaUJs of Hadronic Matter, ed.
by A. Zichichi (Academic Press, New York and London, 1975).
See also Ref. 6.
11. And sometimes given by me. I thank Arthur Wightman for awaken-
ing me from my dogmatic slumbers.
12. It suffices to assume that the gauge field is without (gauge-
invariant) singularities if we make a stereographic projection
of four-space onto a four-sphere. I would love to found the
analysis on finiteness of the action, without even this as-
sumption about the behavior of the fields at infinity, but I
have not been able to do so.
13. Sign convention: In n-space, E: = 1. Symbols with upper
1 n
indices are defined by raising with the appropriate (Euclidean
or Minkowskian) metric.
14. R. Bott, Bull. Soc. Math. France~, 251 (1956).
15. See Ref. 10.
16. See Ref. 6.
17. At least in a box; see the next paragraph.
18. R. Jackiw and C. Rebbi, Phys. Rev. D 14, 517 (1976).
19. Even this is true for axial gauge (in infinite space) only if
we add additional gauge conditions. (See Ref. 6.)
20. M. Atiyah and R. Ward, Comm. Math. Phys. ~, 117 (1977).
21. Although I have just argued that this knowledge is irrelevant
to our immediate purposes, an enormous amount has been learned
recently about solutions to the Euclidean gauge-field equations.

In fact, "binstantons" don't exist, but S!vl-parameter families
of solutions with winding number V do. For a review (with
references to the original literature) see R. Jackiw, C. Nohl,
and C. Rebbi, "Classical and Semiclassical Solutions to Yang-
Mills Theory" (to appear in the proceedings of the 1977 Banff
School, to be published by PlenuIT. Press).
22. For a review of the renormalization group applied to gauge
theories, see Ref. 10.
23. Done by 't Hooft, a hard worker (second paper cited in Ref. 9).
't Hooft's computation has been somewhat simplified. See
A. Belavin and A. M. Po lyakov , Nucl. Phys. B 123, 429 (1977);
F. Ore, Phys. Rev. D16, 2577 (1977); S. Chahda, A. D'Adda,
P. di Vecchia, and F. Nicodemi, Phys. Lett. 72B, 103 (1977).
24. The analysis reported here is based on C. Callan, R. Dashen,
and D. Gross, Phys. Lett. 66B, 375 (1977). The fact that the
Abelian Higgs model in two dimensions doesn't display the
Higgs phenomenon was discovered independently by two of my
graduate students, Frank De Luccia and Paul Steinhardt. They
didn't write up their results because I didn't believe them.
I take this occasion to apologize to them for my stupidity.
25. The problem is identical to that of constructing flux tubes in
superconductors. See Ref. 6, and references cited therein.
26. Indeed, 8 vacua, with precisely the same interpretation (but
derived in a completely different way), occur in the massive
Schwinger model, quantum electrodynamics of charged fermions
in 1 +1 dimensions. [See S. Coleman, R. Jackiw, and L. Suss-
kind, Ann. Phys. (N.Y.) 21, 267 (1975), and S. Coleman, Ann.
Phys. (N.Y.) 101, 239 (1976).] The arguments that work for
the Schwinger model also work for the Higgs model when ~z is
positive, so we also obtain 8 vacua in this case despite the
absence of instantons.
914 s. COLEMAN
27. K. Wilson, Phys. Rev. D10, 2445 (1974). The standard expres-
sion for H has a factor of -iq where I have one of -q/e; the
source of the difference is the factor of ie hidden in my
definition of ~.
28. S. Weinberg, Phys. Rev. Dg, 3583 (1975). This paper, titled
"The U(l) Problem", gives a characteristically lucid descrip-
tion of the situation just before 't Hooft's breakthrough.
(As a major unanswered question, Weinberg lists "How does the
underlying gluon-gauge invariance enforce the equal coupling
of the positive- and negative-metric Goldstone bosons to gauge-
invariant operators?")
29. S. L. Adler, Phys. Rev. 177, 2426 (1969). J. S. Bell and R.
Jackiw, Nuovo Cimento 60, 47 (1969). W. Bardeen, Phys. Rev.
184, 1848 (1969).
30. J. Kogut and L. Susskind, Phys. Rev. D1l., 3594 (1976).
31. For more details on Fermi integration, see F. A. Berezin, The
Method of Second Quantization (Academic Press, New York and
London, 1966).
32. The easiest way to fix this up is to stereographically project
Euclidean four-space onto a four-sphere; is then projected
into an operator with a pure discrete spectrum. This changes
the determinant, but only by a factor that is independent of
the gauge field. Since, as we shall see, our final results
will only depend on ratios of determinants, this change is
irrelevant.
33. To my knowledge, this sum rule was first derived by A. S.
Schwarz, Phys. Lett. 67B, 172 (1977). The derivation in the
literature closest to the one given here is that of L. Brown,
R. Carlitz, and C. Lee, Phys. Rev. D~, 417 (1977).
34. This section is mainly afterthoughts; I didn't know most of
these things at the time these lectures were given.
35. A related picture of how instantons break SU(2) ~ SU(2) is
advanced by D. Caldi, Phys. Rev. Lett. ~, 121 (1977).
36. An (apparently) very different picture of how merons effect

confinement has been advanced by J. Glimm and A. Jaffe ["A
Droplet Model for Quark Confinement" (unpublished)].
G. 't Hooft has advocated completely different configurations
["On the Phase Transition Towards Permanent Quark Confinement"
(unpublished) ] .
37. The prededing paragraph is the product of conversations with
Michael Peskin, who has observed that a group of two-dimensional
models analyzed by C. Callan, R. Dashen, and D. Gross [Phys.
Rev. D~, 2526 (1977)] display (in a certain sense) a restora-
tion of chiral symmetry at large scales, the first half of the
above scenario.
38. R. Crewther, Phys. Lett. 70B, 349 (1977).
39. R. D. Peccei and H. R. Quinn, Phys. Rev. Lett. ~, 1440 (1977);
Phys. Rev.D~, 1791 (1977). F. Hilczek, Phys. Rev. Lett. 40,
279 (1978). S. Vleinberg, Phys. Rev. Lett. ~, 223 (1978) and
"Instantons Hithout Axions" (unpublished).
40. These ideas are the product of discussions with S. Glashowand
D. Nanopoulos.
41. The key paper on this subject is M. B. Voloshin, I. Yu.
Kobzarev, and L. B. Okun, Yad. Fiz. lQ, 1229 (1974) [Sov. J.
Nucl. Phys. lQ, 644 (1975)]. The instanton approach to the
problem was developed in S. Coleman, Phys. Rev. D12, 2929
(1977), and C. Callan and S. Coleman, Ref. 8. (Large portions
of the text of this section are plagiarized from these two
papers.) Similar ideas were developed independently by
M. Stone, Phys. Rev. D14, 3568 (1976) and Phys. Lett. 67B, 186
(1977). P. Frampton was the first to study these phenomena in
the lVeinberg-Salam model [Phys. Rev. Lett. ]2, 1378 (1976)];
however, Frampton's conclusions have been criticized severely
(and, I think, correctly) by A. Linde (unpublished).
42. S. Coleman, V. Glaser, and A. Martin, Comm. Math. Phys. (in
press).
916 S. COLEMAN
43. See, for example, R. Jackiw, Phys. Rev. D2., 1686 (1974).
44. Formulas related to the one developed here can be found
throughout the literature. Two references out of many:
J. H. Van Vleck, Proc. Nat. Acad. Sci. 14, 178 (1928).
R. Dashen, B. Hasslacher, and A. Neveu, Phys. Rev. D 10, 4114
(1974). The derivation given here was developed in conversa-
tions with Ian Affleck.
45. This appendix reports on computations done with C. Callan. A
somewhat different attack on the problem (with the same con-
clusions) is E. Gildener and A. Patrascioiu, Phys. Rev. D16,
423 (1977).
46. I give no reference not because these results are novel but
because they are a standard part of the theory of Weiner in-
tegrals.
47. This appendix is a transcription of an argument of Brown et
al. (Ref. 33).
ACKNOWLEDGMENTS
After I gave these lectures at Erice, I repeated them at the

Ecole Normale Superieure in August and the Lawrence Berkeley Labor-
atory in September. I thank these institutions for their hospi-
tality.
During these reruns, I made improvements in parts of the
lectures, in large measure in response to comments made at Erice,
and I have incorporated these in my notes. Thus, if in the tran-
scripts of the discussion sessions you find students asking
questions already answered in detail in the notes, it's not that
they didn't listen, but that I did.
This work was supported in part by the National Science Found-
ation under Grant No. PHY77-22864.
DIS C U S S ION
CHAIRMAN: Prof. S. Coleman

Scientific Secretary: B.M. Schechter
DISCUSSION No.1
BUNK:
In the example of the double well potential you

integrated over all n-instanton configurations. What is
the justification for this procedure insofar as there is
no topological distinction between them? I feel there is
a danger of double counting.
COLEMAN:
Well, I'm certainly not double counting. Don't

forget, I'm only integrating over an immediate neighbour-
hood of each of the n-instanton configurations. I'm
doing a Gaussian integration and the Gaussian falls off
quite rapidly in all directions as you move away from
these configurations. So I'm certainly not double count-
ing. My approximation for the integrand goes to an
incredibly tiny number before I reach another one of
these configurations.
BUNK:
But you also had to integrate over instanton posit-

ions and there it occurred that an instanton and an anti-
instanton could be quite close together.
917
918 DISCUSSION
COLEMAN:
That's right. But that's a very small fraction of

the space. I'm totally neglecting interactions between
the instantons. In principle I should cut off some
arbitrarily chosen but quite sizable region where the
instantons overlap. The result of that is negligible.
In statistical mechanics this sort of an approximation
is known as the dilute gas approximation.
PAFFUTI:
The mUlti-instanton in the example of the double

well potential are not solutions of the equation of
motion so why do you consider them?
COLEMAN:
Well, it's very close to being a stationary point.

When the instantons are far apart the change in the
action when I change the inter-instanton distance is
very tiny. If I could in a sensible way speak of a
string of instantons with infinite separation between
them <of course, I can't) that would be a stationary
point of the action. As time gets very, very large they
do not have an infinite separation but they do have a
large separation, so it's very close to being a station-
ary point of the action.
Consider the example of the integral of a function
like e~~GS~)where S looks like
)(
It goes asymptotically to zero at infinity. It's obvious

where the dominant part of this integral is going to come
from: the region out towards infinity; the stationary
point is at infinity. And this becomes more pronounced
as ~ goes to zero. That's what things are like for the
multi-instanton case with x corresponding to the instanton
separation. There's a stationary point at infinity and
you have to include that in the integral, otherwise you
will make errors. As this example shows, you are liable
to get dominant contributions to an integral of this kind
DISCUSSION 919
not only from finite parts of the plane where the inte-
grand is stationary, but if the derivative of the inte-
grand goes to zero as its variable goes off to infinity
then you'd better also worry about infinity.
TOWNSEND:
In the example of the double well potential you
computed the transition amplitude from +a to -a. This
required considering chains containing even numbers of
instantons. Had you commuted the amplitude for transit-
ions from -a to a you presumably would have had to
consider chains containing odd numbers of instantons.
Would you comment on how this changes the results?
COLEMAN:
You get, of course, exactly the same sort of thing.
Instead of the series for the sinh you get the cosh, but
still you get an appropriately weighted sum of two ex-
ponentials. And of course you get the right answer. In
this case you get
..... 1 (~ ):'i [E+T/i\
<
-q, t e
-+It:
\+(ll' = -
2.
-
11"'-
e - e'f..T/jt]
with a minus sign this time since the odd terms of the
series contribute instead of the even ones. The even
eigenstates contribute with the plus sign and the odd
with a minus sign. So, you know, whatever you want to
compute is fine: you always get the right answers. You
have to, the method is right.
MOTTOLA:
The n-instanton configuration functional integral
involves a determinant which is approximately the single
instanton determinant raised to the nth power. Is there
an easy way to see this?
COLEMAN:
Yes! I'll begin by making some definitions:
920 DISCUSSION
then we can write the determinant as a quotient
'Z ~
det--X2.t-(jt'+~+tt
"2. (
~-Tk
)
]
de~-'Y2. [- '?>~ +(A)"&]
where Tk is the position of the kth instanton. Using

the famous formula for the determinant
I obtain n..
= e){~ {_ ~ T'C' ..en l1 + ~ _~::~'Z f \+.- T" )) \ .
We all know how to evaluate this - it's just Feynman's
( "" ..... "-,1
perturbation theory with -"t+(A)) as the propagator
and f as an external source. So this is just the sum
of all the connected graphs
Tz, "l1 1
x----O---x + etc.. S .
If the Tn's are widely separated there's absolutely no

chance of there being a cross term between two different
Tn's because the propagator is falling off. So you obtain
n
de\:-'Yz [- '?>~ + ~ilt:-TJc.)+(b"Z.J = det:-~ [_~:;o(J)1.+ of (t) J
det:-y~ [- ~~ +fi] dQt-~ [-~t+~]
But the determinant you get from n-instanton configurat-

ions is not the determinant you get from a single ins tan-
ton raised to the nth power. You have to divide out the
region in between the instantons before you can conduct
the argument.
DISCUSSION 921
DISCUSSION No.2 (Scientific Secretary: F.J. De Luccia)
BOHR:
What is the difference between the Pontriagin index
and the winding number?
COLEMAN:
The Pontriagin index is just what I called the wind-
ing number. I called it the winding number in these
lectures just because I didn't want to introduce a lot
of terminology.
BOHR:
What can you do if your Lie group is not compact?
COLEMAN:
You mean if it's not simple. All gauge groups are
compact, so for our purposes we need only consider a
compact group. A compact group is locally a product of
factors - and it turns out to be only the local structure
that is relevant here - the product of a bunch of U(l)'s
and simple groups. For each U(l) there is no contribut-
ion to the Pontriagin index or winding number - no effect.
Every mapping of S3 to the circle can be continuously
deformed to the trivial mapping. For each simple group
in the product you have a winding number, and that wind-
ing number is in fact always computed the same way, by
the formula:
on:. __
1
1E.1t'2
5dx+ k- l FF)
,..,
.
If you have normalized the group matrices as I have done

in the lecture, this is the universal formula for the
winding number. The reason it is the universal formula
is that there is a theorem of Raoul Bott that says that
if you have any mapping of a 3-sphere into any simple
Lie group you can continuously distort it completely
into an SU(2) subgroup. So every simple group is the
same as SU(2) as far as computing the Pontriagin index
922 DISCUSSION
goes. I'm making no simplification whatsoever when I

restr ict mys elf to SU ( 2 ).
WEEKS:
In classical Yang-Mills theory do you include in
your gauge group the transformations with winding number
different from zero or those that are disconnected from
the identity.
COLEMAN:
The full gauge group of the theory is the set of
functions from all of 4-space into the group. There is
no winding number attached to these mappings.
WEEKS:
In QED if you don't make some boundary condition at
infinity you find out that the Coulomb gauge is not a
gauge because the Laplacian is not invertible.
COLEMAN:
Physics determines the boundary conditions at infin-
ity. There's a physical reason why in QED when we work
in the Coulomb gauge we throwaway the solutions of the
Laplace equation that grow at infinity. Let me consider
the simplest case. Suppose the potential is a constant
times x, so that there is a constant electric field in
the x-direction. This is a perfectly legitimate solution
of Laplace's equation, and if you wish you can put it at
the beginning in QED. You can imagine someone has erect-
ed giant condenser plates at opposite ends of the galaxy
and charged them up so that there is a constant electric
field permeating all of space. Would you get a different
QED than you would get if you assumed that the electric
field goes to zero at infinity? The answer is no, becaus4
as soon as you couple in the electrons you make the vac-
uum polarizable, and the vacuum suffers dielectric break-
down. Electron-positron pairs will be pulled out of the
vacuum, and if the condenser plates are far apart, they
get a lot of energy as they travel t'o the plates, more
than enough to compensate for the 2mc 2 required to make
them. The electrons would fly to the condenser plates
until they were neutralized; and, after this enormous
DISCUSSION 923
explosion, you would end up with exactly the same theory

you find described in Bjorken and Drell. This is why
you may as well assume the fields go to zero at infinity.
JANCEWICZ:
I did not understand yourexample with the two gauge
fields with the same winding number. You said that these
two fields are connected by a gauge transformation, but
the gauge transformation you chose is the identity at
the origin.
COLEMAN:
No. Let's remember what the theorem was. I stated

that there is a gauge transformati~n U(x) defined on all
sRace such that when applied to A~" it reduces it to
A!) at infinity. I didn't say they were gauge equival-
ent everywhere - that would be obviously preposterous.
DISCUSSION No.3 (Scientific Secretaries: G. Paffuti,

J. Preskill)
PAFFUTI:
The coupling constant which you use in the express-

ion for tunnelling is the effective coupling constant,
which can be computed by renormalization group techniq-
ues. In this case, what does a ~- function calculated
in a perturbative way mean, if the vacuum of the theory
is not the perturbative vacuum?
COLEMAN:
Firstly, although it is not perturbative, this is
a systematic weak coupling expansion. Thus, all argu-
ments you make about the orders of things are the same
as you would make in perturbation theory. In particular,
all renormalizations occur just as in perturbation theory.
For example, though the determinant K is formally ultra-
violet divergent, that divergence is cancelled by the
standard counterterms, and the finite parts of those
counterterms are determined by the renormalization con-
ditions. There is reference to the renormalization
924 DISCUSSION
point in the determinant K.

When you carry out the calculation of the determin-
ant, you know what is going to happen: The one instan-
ton contribution to the functional integral, for an
instanton of a given size ~ , must be of the form
d\AI..ln(1t2 ').2.) _9-
jc.' e r1 '-I"' e ~2.
where ~1 is the usual renormalization group coefficient
and ~~s the renormalization point. Otherwise it is not
renormalization group invariant. The combination ~'ZA.z.
must occur by dimensional analysis; there are only two
lengths in the problem. The effect of this is that it
is the running coupling constant which occurs in the
expression. This factor suppresses the contributions of
the tiny instantons, but it doesn't help with the fat
instantons. You need new physics there.
PAFFUTI:
How good an approximation is it to compute the eff-
ect of tunnelling with a gas of noninteracting instantons?
COLEMAN:
Very good, as long as the coupling constant is very
small, and there are no long range forces between the
instantons. Suppose the instantons have a definite size,
so we can ignore the integration over instanton scale.
Then we are doing a sum of the form
_ sol1\. \Y\
L.!.le KVT)
n. n!
The dominant contribution to the sum comes from the terms
for which
01'
We see that as long as ~ is small, the density of in-

stantons is a very small number, and the corrections to
the dilute gas approximation go as the square of the
density, which is even smaller.
DISCUSSION 925
KENWAY:
This morning you showed how instantons can restore

broken symmetry leading to confinement in two-dimension-
al theories. Callan, Dashen, and Gross suggest that in-
stantons, as such, are not responsible for confinement
in four dimensions, but that there may be a phase trans-
i tion at ~ '" 1 into a plasma of merons, which are obj ects
of fractional winding number. In this phase color may
be confined. Could you explain this? Specifically, what
is the relevance of field configurations whose action is
not a minimum within a homotopy class?
COLEMAN:
I have no profound insight into their computation.
I look at it and say, "Who would believe that?" But, on
the other hand, I say, "Is it evident nonsense?" I prefer
to say nothing at all about it here.
As for the fact that instantons will not give rise
to confinement in four dimensions, that's obvious. We
do the loop calculation again. Instantons near the
boundary of the loop can give rise to a large effect, but
they contribute a factor like ~X~(L+'). That doesn't give
us a linear potential. We need a contribution from the
instantons inside the loop. We might naively expect to
get one, because the gauge fields far from an instanton
falloff like l/r. But the part of the instanton that
goes as l/r is a pure gauge, and, in four dimensions, I
can make a gauge transformation such that the gauge field
vanishes outside a small tube perpendicular to the loop,
so it doesn't contribute at all. There is no hope of
getting a contribution from the instantons that goes as
e)(p- c..A~e.o. ') .
't HOOFT:
Polyakov succeeded ~n doing such a thing in three

dimensions.
COLEMAN :
But that's completely different. Polyakov does not
use the dilute gas approximation. He studies the Georgi-
Glashow model, in which the instantons are the
926 DISCUSSION
't Hooft-Polyakov monopoles. There the instantons do

have a long range interaction - the Coulomb interaction.
Polyakov goes beyond the dilute gas approximation, but
his tricks will not work in four dimensions.
WIGHTMAN:
Is it not true that the probability for a field to
approach infinity at r = 00 is finite when measured in
the probability measure defined by the action? (In fact,
I know that for a free scalar field the answer is yes;
see the article by Lanford and Collela in the 1973 Erice
Mathematical Physics School.) Thus, the probability
that the action is finite is zero. It is remarkable
that in spite of this, the points in function space where
the action is finite and stationary seem to dominate the
integral. Do you have any comment on this situation?
In particular, if one writes
r -~/-n.
J [dA1 e. 0\- " 'Rest"
n-\'ilted
S -ti~ite
do you have any argument why "Rest" =0 ?
COLEMAN:
I know of no such argument. But the problems arise
just as much in one-dimensional quantum mechanics as they
do in field theories. If they cause a disaster, they
should cause a disaster in the potential problem consid-
ered in the first lecture. But they don't. You get
exactly the same answer using the functional integration
methods as in good old wave mechanics using WKB.
Also, when you look at one-loop corrections in four
dimensions, of course, you get infinities, but they are
cancelled by the usual renormalization counterterms. The
renormalizations work out exactly right using the funct-
ional integration. So, at the level, not of a quantum
field theorist, but of a quantum field butcher, the level
at which I operate, so far, I would say, things look good.
't HOOFT:
I have one remark. In four dimensional gauge

theories, your determinant K goes like a large negative
DISCUSSION 927
power of g, namely ~-s for SU(2) and ~-12 for SU(3).

If you treat g as a running coupling constant g (\-I.)
here, and assume that for }J. small, g -+ 00 then you
get a nice natural cutoff for your integration over the
size parameter of the instanton. Of course, this is not
qui te legal because as g -+ 0() the approximations make
no sense, but maybe it is physically reasonable.
COLEMAN:
Yes. One can observe these negative powers in my
own treatment of the potential problem. Remember that
I found a factor of r.
Soli\.1~ which came from the change
of coordinates that eliminated the translation mode. In
four dimensions, this would be raised to the fourth power,
and additional powers come from zero modes associated
with rotations in the gauge group, as you point out. I
have no comment, except that this is a clever idea about
the source of the infrared cutoff.
't HOOFT:
You consider the 1+1 dimensional model in the Higgs

mode and find that nevertheless instantons provide con-
finement. Does this also change the particle spectrum?
You had a massive vector and a massive scalar.
COLEMAN:
I don't know. I do expect there to be charged part-
icles in the theory because of the [1- Cd..l21Tq/e)] factor
in the force. This vanishes for q an integral multiple
of e, which I can only explain by the mechanism I ment-
ioned, the breaking of the string. So there are part-
icles of charge e which are confined in a linear pot-
ential which is very shallow, because of the ~~GS~~)
factor. Hence, I expect a large number of closely spaced
neutral bound states, as in the massive Schwinger model.
Whether some of these bound states can be identified with
the Higgs meson or the massive vector meson, I don't
know. I do know that this rich spectrum of bound states
is not what I would get by applying perturbation theory
to the Abelian Higgs model.
928 DISCUSSION
't HOOFT:
Then my last question is probably a rhetorical one.

It seems that the Higgs mode then does not differ essent-
ially from the symmetric mode. Do you expect a phase
transition between the two or not?
COLEMAN :
I think there is a phase transition, because there

is an essential singularity in e~ in the Higgs mode, but
I am not sure.
TOWNSEND:
You have shown that in a two-dimensional model, the

instantons restore the symmetry. Isn't this to be ex-
pected from your theorem which says that a continuous
symmetry cannot be spontaneously broken in two dimensions?
COLEMAN:
There is a theorem due to me which says you cannot

have the Goldstone phenomenon in two dimensions. This
is not the statement that you can have no Higgs phenom-
enon in two dimensions. The conventional Schwinger model,
with massless fermions, does, as shown by Lowenstein and
Swieca, act just like a Higgs model with respect to the
charged current when you quantize it in a covariant gauge.
So the Higgs phenomenon can exist in two dimensions. But
what does "Higgs phenomenon" mean? I like to say that
the Higgs phenomenon is an explanation, not a phenomenon.
It is not like Goldstone. The Goldstone phenomenon means
there is a conserved current and the associated charge
does not annihilate the vacuum. The Higgs phenomenon is
not so easy to get a handle on. Let me give you a simple
but very instructive example.
In four dimensions, consider
= i l?>.. cPJ
2.
~
the free massless scalar field theory. It has the invar-
iance
DISCUSSION 929
and the associated conserved current

'J"p. =- O!Je\>
The Goldstone phenomenon occurs, in that the symmetry
is not preserved by the vacuum: < 4> > is indeed changed
when I add A to it. I now Higgs the model:
cf = _.1 (f~" )2. -t ( {)~ + A}I )2-

4-
This is the simplest Higgs model. If we define
then
~ = -1 (F;" y~ + B~ .
I now have a theory of a massive vector meson. But no
one presented with this final theory could possibly tell
it was a result of the Higgs phenomenon; it's just the
theory of a free massive vector meson. Unlike the
Goldstone phenomenon, the Higgs phenomenon leaves no
footprints. In fact, in simple models that you can anal-
yze, it's very easy to prove that in the gauge Ao=o
the vacuum expectation value of the scalar field is zero.
In a covariant gauge, of course, it can have any value
around the circle. So it's a bad way to ask the question
to say~"Do the instantons destroy the Higgs phenomenon?"
You can destroy the Higgs phenomenon by adopting a diff-
erent gauge.
JANCEWICZ:
In two-dimensional spacetime you have only one comp-

onent of the electromagnetic field; it is the electric
field, of course. Why do you call the curvilinear integr-
al a magnetic flux?
COLEMAN :
In Euclidean space, all dimensions are spatial, so

all components of ~~ are space-space components and
hence "magnetic" in character. But I'm just making an
analogy when I speak of magnetic flux. There is no flux
quantization back in Minkowski space.
930 DISCUSSION
DISCUSSION No.4 (Scientific Secretaries: J. Preskill,

B. Schechter)
PAFFUTI:
You can subtract a topological current from the

current y~S to obtain a conserved current. Why doesn't
the Goldstone theorem apply? Is it because the con-
served current is not gauge-invariant?
COLEMAN:
Yes, let me explain what happens. We have a current

:filS such that
in the case where there is only one flavor. But there is

another current
(\...5
O~ = <bll'~
"
f: ",~').c:r T'C" \..A-J FACT - i A~ A).Ac:r)
which, although it is not gauge-invariant, obeys
~ ~ c9:s = ~ T<" Cf' f)
/.I. \1t z
in any gauge. Therefore we can construct a conserved but
gauge-var~an. t curren t Ie:. f.l = :r:~S - (\.5
e". .
Now the standard Goldstone theorem should apply. In part-
icular the ""5- symmetry multiplies G'" -= ~ t 1+'1'.') '-\I by a
gauge, <
\C/.I.(~) <r(o) >
phase factor, so the theorem states that, in a covariant
must have a Goldstone pole; that
is, a pole at zero mass in its Fourier transform. Indeed,
if we do a sum over instantons to compute it, we find that
<
k"C.x.) crlD) does have such a pole. However, in the
same gauge, {<rtX.')()(D' and <k... Cx.) lC.olo)) have no
Goldstone pole. The first statement tells us that there
is a zero mass particle which ( j makes from the vacuum
and K" annihilates. The other statements seem to say
that this is not so. What's wrong?
The problem is that we are working in a covariant
gauge, where there are lots of negative norm states. To
see how precisely this structure can arise, consider a
theory in which I have two mass zero scalar fields
r----,
4>lx} cP(o) :: ~ (Fourier transform)

r---"'I
q,'(x)4>'lo)=_l.. (Fourier transform)
\C.1.
DISCUSSION 931
The field ~ is perfectly normal, but ~' has negative

norm, .like a longitudinal photon in Feynman gauge. Now
suppose that <r couples to cP+c\>' and \::of-'. couples to
q,_~' . If I look for a pole in < a-(X) <r(o)
or <I:",(:'lt} )c::olo) there isn't any; the two terms can-
cel. But in <"fl(x) cr(o there is a pole. This
phenomenon, called a Goldstone dipole, was observed by
Kogut and Susskind in the Schwinger model in a covariant
gauge. It explains how the proof of the Goldstone theor-
em can apply to < "",(x) <r(o' but the Goldstone boson
does not matter. It is just a gauge artifact. Gauge-
invariant obj ects like <crlx) ~lo)) have no poles in them.
LIPKIN:
Hadrons now have four, five, or more flavors, but

Goldstone bosons have only two. Does this mean that the
u and d flavors are more equal than the others? Also,
what is the ~ and is it accidental that the ~ and
K masses fit the Gell-Mann Okubo formula together with
the pion, which is a Goldstone boson?
COLEMAN:
The answer is the old hierarchical scheme. SU(3)

is a pretty good approximation, and chiral SU(2) X SU(2)
is a very good approximation. If you combine them, as
is typically the case when you combine a very good
approximation with a pretty good approximation, chiral
SU(3) X SU(3) is at best so-so. The question is, what
limit are you talking about? In the chiral SU(2) X SU(2)
limit, the pions are Goldstone bosons, and the ~ and
K are just some particles about which chiral SU(2) X
SU(2) tells you nothing. In the less realistic chiral
SU (3) X SU (3) limit, the 1(. 's and K' sand 'Yl. are all
massless Goldstone bosons. You still have a U(l) problem,
but it has nothing to do with the ~ which is a Goldstone
boson of chiral SU(3) X SU(3). The object associated
with chiral U(l) is a particle with the quantum numbers
of the 'l. which is not there. It is not something in
the Rosenfeld table, and has nothing to do with the~
in any limit.
It is clear that the pion can be both an almost-
Goldstone boson and part of an SU(3) multiplet with the
~ and the K's. There is no contradiction in these
statements. It is not clear why the Gell-Mann-Okubo
932 DISCUSSION
formula works so well for the pseudo-scalar meson octet,

where the mass splittings are so large. A variety of
explanations have been proposed, but I think there is no
general agreement.
't HOOFT:
Might it not make sense to consider a "skeleton"

theory without instantons, which would have four
Goldstone bosons? Then if you switch on the instantons,
you probably produce an 'f\,-rt: mass matrix.
COLEMAN:
That depends on what you think about the dynamical

breaking of SU(2) X SU(2). First of all, I am not at all
sure that it is sensible to think of the theory as if
there were no instantons. If it is a sane thing to do,
I can consider three possibilities. The first possibil-
ity is that both chiral SU(2) X SU(2) and chiral U(l)
break spontaneously even in the absence of instantons,
so that there are four Goldstone bosons. Then the inst-
antons would come along and shove up the mass of one of
them, just as you suggest. A second possibility, however,
is that the dynamical mechanism of qeD which is respons-
ible for the breaking of SU(2) X SU(2) is driven by the
instantons in some indirect way. Then, when we neglect
instantons, there is no symmetry breaking, and there is
never any massless ~. The third possibility is that
when the instantons are absent, chiral SU(2) X SU(2) is
spontaneously broken, but chiral U(l) remains as a good
symmetry. When you turn on the instantons, they break
chiral U(l), but again, you cannot think of the theory
as containing a massless ~ in any limit. At the current
state of our knowledge, I, at least, cannot see any rea-
son why any of these possibilities should be excluded.
This is a question that would much profit from further
investigation, and I encourage people to investigate it.
PRESKILL:
You argue that instanton effects give rise to a

confining linear potential in the two-dimensional Abelian
Higgs model. You quoted an expression for the form of
the potential in the e =0 case, which contained the
factor t 1- Co~ (21tq/e) '] If one does the computation
DISCUSSION 933
for arbitrary e this factor is replaced by

[cesS - (0'0. (.211q/e + e)]
Now one notices that this expression, with, for example,
~ =It has a different sign than for e = 0 This indic-
ates that, in some of the vacua, we have, instead of
confinement, a linear repulsive interaction between
quarks. Perhaps this has a background field effect of
some kind. Could you comment?
COLEMAN:
Yes, that is the answer. Expand the expression for

.Q and 21r~ /e small:
E cC ~ L ( 2~q + e') e
2
_ To ]
Now let's consider a one-dimensional world in which I

have a constant electric field G.
Suppose I introduce a
pair of charges :t 9 separated by a distance L. In the
weak field approximation, where I neglect vacuum polar-
ization effects, the field between the charges is now
t+q. Now I consider the energy difference between the
configuration with the external charges and that without.
The energy density is proportional to the square of the
electric field, which is the same in both cases except
for a region of length L.
E OC !::
2
I <. ~+~ )2_ c'l.] .
I propose to you that there is a similarity between these
two expressions, if you identify ~ with a mUltiple of
the external electric field ~. You can easily see,
therefore, why you sometimes get repulsive interactions,
and also why e
is a periodic variable. If 1'0 \ > e/2. )
then, by materializing a pair of charges ~e from the
vacuum, I can reduce the energy. Pairs will continue to
form until - e/2 ~ C ~ e/2. The magnitude of the field
between the charges can then be either greater than or
less than the background field, so the interaction is
sometimes attractive and sometimes repUlsive.
TOWNSEND:
If you couple an axial gauge field to the chiral

current does this gauge field become massive when chiral
symmetry is spontaneously broken?
934 DISCUSSION
COLEMAN:
Because there is an anomaly, it makes no sense to

think of "gauging" the U(1) current; you can only
"gauge" conserved currents.
CELMASTER:
You said that in order that the action be finite

the field A must be given at infinity by
A~t~)::: U(x) ~\1 U-~\X) -+ 0(1/,.2) .
Can you in fact demonstrate that A~ becomes dependent
only on direction?
COLEMAN:
This fact, while tacitly assumed, is actually not

needed. All we need is that A~ go to a pure gauge for
R~ sufficiently large. Then on the surface at R we
may define a winding number which remains unchanged as
we continuously enlarge R. It must. Expansion is a
smooth operation which keeps us within the same homotopy
~lass. This is really the only fact that we needed.
It HOOFT:
In your nice demonstration of how the Adler-Bell-

Jackiw anomaly equation gives us the number of zero
"energy" modes, one feature seems to stay somewhat obsc-
ure. The anomaly was always thought to come about from
from an ambiguity in the renormalization procedure: If
you regularize one way you spoil gauge invariance; the
other way you spoil chiral symmetry. Now suddenly you
deri ve a zero "energyl1 state without ever discussing
high frequencies. Is this not worrying you?
COLEMAN :
To hell with where it comes from! I integrate the

anomaly equation over space-time: there is no better
way of obtaining a zero frequency phenomenon. If the
Adler-Bell-Jackiw equation is correct it has to give the
correct answer when applied here.
DISCUSSION 935
DISCUSSION No.5 (Scientific Secretaries:

F.J. De Luccia, G. Paffuti)
CABIBBO:
Assuming we are living in a false vacuum, are there

some things we should not do?
COLEMAN:
The question is: can we artificially induce vacuum

decay, assuming that the probability of the vacuum de-
caying spontaneously in the lifetime of the universe is
negligibly small? I can't compute the probability of
induced decay, but there is nothing to worry about with
respect to the next generation of high-energy acceler-
ators because there are very high energy proton-proton
collisions going on right now when cosmic rays hit plan-
ets or stars, and these energies are much higher than
anything we can produce in the laboratory. There may
not be many of these events occurring in our immediate
vicinity, but remember that if a bubble had formed any-
where in our past light cone, we would have seen the
effects. So even though these very high energy events
may be quite rare, there probably have been many more
of these events occurring in our past light cone than the
human race will produce in the laboratory in the fore-
seeable future. So at a practical level this is not a
serious worry.
Now how do you estimate the probability of induced
decay? It's a difficult problem, but I have a qualitat-
ive argument that shows that it is smaller than you might
think. If you look at a proton-proton collision in the
center-of-mass frame, you can easily build up enough
energy per unit volume to make a transition. However,
very high energy protons are represented by very high
frequency waves - they are wiggling up and down a lot -
but the bubble is not an object with a lot of wiggles in
it. What you need to make a bubble, I think, and this
is just guesswork, is a coherent state. You need a super
laser that makes the amount of energy you have in cosmic
rays, but with a lot of low energy mesons or something.
That's what is going to push the vacuum over smoothly.
936 DISCUSSION
WHITE:
Could you suggest what experimenters can do to test

the validity of quantum chromodynamics?
COLEMAN:
The first question one should ask is why theorists

believe in QCD. Part of it is fashion following. Gauge
theories have scored a brillant success in weak inter-
action physics, so why not apply them to the strong inter-
actions? Secondly, there was the discovery of asymptotic
freedom, that there is a sensible way of explaining
Bjorken scaling, or something closely approximating it,
in a quantum field theory. This makes a gauge theory of
the strong interactions look very, very good. Thirdly,
there are many handwaving ideas that no one can show are
rigorous consequences of QCD, but which are consistent
with the ideas of QCD, things like quark-antiquark bound
states, three quark bound states, but no quark-quark
bound states, and so on. So there are two classes of
predictions I'd like to distinguish between, one which
consists of rigorous consequences of QCD like asymptotic
freedom, the other which consists of handwaving ideas
like quark confinement. The rigorous predictions are
hard to test. They require high precision measurements
of the moments of the structure functions of deep-inel-
astic scattering processes. This won't be done within
the next few years. But if eventually these measurements
are made with sufficient precision and the logarithmic
factors predicted by asymptotic freedom are not there,
then, as far as I am concerned, that kills the theory.
That's a very sharp test. You can also attack the hand-
waving ideas, but then the QCD theorist can say you
haven't proved anything since these are not really rig-
orous consequences of the theory.
I think the most striking application of the rough,
handwaving ideas of QCD is in the phenomenology of the
j/~ system. One takes the naive quark model, the
Coulomb potential from asymptotic freedom, the linearly
rising potential from Ken Wilson's work on lattice gauge
theories, one puts these ideas together, and one gets
pretty good predictions, apart from a few embarrassing
points, such as some branching ratios, and so on. These
are all handwaving ideas, but they are characteristic of
QCD. SO the predictions are pretty convincing and
DISCUSSION 937
attractive. The trouble is the predictions were made

after the discovery of the ~/~system. However, if
Lederman's particle is just the J/~ system allover again,
then we've got a real crunch, because now the theories
are established for the ~/~ system. For this new system
all one would have to do is adjust the mass of the quark
and do the same calculations - no further cheating allow-
ed. So that's going to be pretty interesting, and unlike
the precision measurements of deep-inelastic structure
functions, that's going to be happening in the next few
years, or maybe sooner.
BUDNY:
Could it be that the big bang was a bubble?
COLEMAN:
Not as I've described it. You have to have a start-

ing point before the bubbles begin to form. It cannot
be infinitely far in the past, or else the bubbles would
have formed infinitely far in the past. The idea is -
and this is not serious physics - that you have some
universe, maybe with some matter in it, maybe not, and
it bubbles. Since everything is happening essentially
at the speed of light, the average distance between
bubbles and the time it takes bubbles to form are of the
same order, because this is the only scale in the theory.
These bubbles will form, grow, and begin to clash. When
bubble walls collide, energy begins to be released in
the form of particles, presumably. Eventually the univ-
erse consists of the remnants of the clashing bubble walls.
The walls just don't go through each other, that much I
know. So this is the matter we see. We live in the
debris of the catastrophe. But how to compute the spectr-
um of that matter, I just don't know.
KENWAY:
You studied the collapse of the false vacuum in a

scalar field theory coupled to the classical gravitation-
al field, and you showed that the energy density in the
region of the bubble wall increases with the radius of
the bubble, so it would appear that gravitational effects
might be very large. Can you comment on this?
938 DISCUSSION
COLEMAN :
Suppose I have a system with energy E and size L.

When do I expect gravity to be important? The answer is
when the gravitational potential energy is of the order
of the energy of the system:
GE'2...... E Of" L N EKE
L
For the system we are considering, E", E L3 where Eo is
the energy density difference between the true and false
vacuums. Therefore, we'd expect gravity to become
important when
1
Lr.I&EL~ OT L N
{(it
If you take a typical microphysical energy difference,
for example, the difference between the energy of the
symmetric and asymmetric vacuums in the Weinberg-Salam
model for some randomly chosen Higgs meson mass, you find
that this number is of the order of one kilometer. So
we do not have to worry about gravitational effects dur-
ing the formation of the bubble, a quantum process, but
during the later stages of the evolution of the bubble
you might expect gravitation to become important. In
fact, gravity is not important in the expansion of the
bubble. There are two arguments to support this, and
Frank De Luccia and I have done an actual computation to
back this up. The first argument is as follows. Assume
for simplicity that the catastrophe has yet to occur, so
that outside the bubble, space is essentially flat. Con-
sider how the expansion of the bubble looks to an obser-
ver outside the bubble wall. I have already argued that
gravitation is unimportant during the early stages of the
bubble expansion, so during this time the bubble wall
describes a hyperboloid. The equations of motion are all
Lorentz invariant - the gravitational field equations do
not break Lorentz invariance. Since the equations are
0(3,1) invariant and the initial conditions are 0(3,1)
invariant, so is the solution, so the expanding bubble
wall is still a hyperboloid. The second argument is to
appeal to Birkhoff's theorem which says that the gravit-
ational field outside a spherically symmetric expanding
or contracting object depends only on the total energy
within it. Now the total energy of the bubble is the
sum of two terms, the positive energy of the bubble wall
and the negative of the interior. As I have shown,
DISCUSSION 939
these two energies cancel, so as seen from outside, the

total energy of the bubble is zero. Therefore, the
gravitational field as seen by a little man riding on
the surface of the bubble wall has no effect, and he
continues moving on a hyperboloid. Now there is a long
range effect. When one solves the coupled scalar field
and gravitational field equations, one finds that the
metric inside the bubble is that of de Sitter space.
Equivalently, there is a nonzero cosmological constant
inside the bubble in this case, in which we are living
in the pre-apocalyptic era.
If, on the other hand we are living in the post-
apocalypic era, that is, the bubble was formed long in
the past, by the same reasoning we must say that before
the bubble formed there was a nonzero cosmological con-
stant, since we know there is none now, and that it was
cancelled out by the change in vacuum energy. We have
de Sitter space outside the bubble and flat space inside.
The moral is that gravity does become important when the
radius of the bubble is Of the order of a kilometer, but
it becomes important for people inside the bubble. For
people outside the bubble, even near the bubble wall,
space is flat.
MOTTOLA:
You gave an argument for the existence of a non-

trivial instanton-like solution which dominates the
tunnelling amplitude in Euclidean space. Please state
the precise necessary conditions for the existence of
this solution.
COLEMAN :
The statement of the theorem is as follows:

Consider the scalar field equation
)'2. q> :: \J'(c.p)
in four dimensions. Without loss of generality I can
assume that the minimum of U(<.p) occurs at q>= c. Suppose
also that UlCP) satisfies the following conditions:
(1) U is continuously differentiable.
(2) U(O) = U'(O) = O.
940 DISCUSSION
(3) U is somewhere negative.

(4) There exist positive numbers a, b, 0. , ~, where
0( < ~ < 4- such that \j _ a. \~\ot+ b \tp \~ ~ O.
Then one can prove:
(1) There is a non-trivial spherically symmetric sol-
ution.
(2) Its action is less than or equal to the action of
any other solution.
(3) If another solution exists and is not spherically
symmetric, its action is strictly greater than the
action of the spherically symmetric solution.
s.,4S
(4) has one and only one negative eigenvalue.
bcp'2.
The proof of these assertions will appear in a
paper by Glaser, Martin, and me.
MOTTOLA:
Is there a converse theorem? Does tunnelling imply

a non-trivial solution of the Euclidean field equations?
COLEMAN:
It is not known whether every process that can be

reasonably called tunnelling can be analyzed by these
methods.
DE LUCCIA:
What guarantees that the subspace of zero-eigenvalue

solutions of the Dirac equation in an external instanton
the indices "+

field is finite-dimensional so that the definition of
makes sense?
COLEMAN:
For the instanton one can show by direct analysis

that it is finite dimensional. I don't know of any gen-
eral argument that says that for any finite action sol-
ution it has to be finite dimensional. Note added:
Wightman tells me that standard methods can-be used to
DISCUSSION 941
show the eigenvalues are only finitely degenerate.
TOWNSEND:
My question concerns the ~ -vacuums in QCD. Is the

value of e physically relevant? If so, then how is its
value actually determined? If it cannot be determined
within QCD, then does this mean that QCD is an incomplete
theory?
COLEMAN:
There are two parts to this question. Is S a com-

pletely arbitrary parameter, and if it is, then how do
we determine it? In the model I described, QCD with one
fermion, it's a completely arbitrary parameter in the
same sense that the ~-model has a completely arbitrary
parameter, the direction in which the ~ -field points.
In more complicated cases, such as the Weinberg-Salam
model with one Higgs doublet, e is physically relevant
(I believe). In other cases, it is not.
Now what do we do when e is physically relevant?
Then it's just another parameter that has to be deter-
mined by experiment, like a coupling constant or the
Cabibbo angle.
CAN WE MAKE SENSE OUT OF "QUANTUM CHROMODYNAMICS"?
G. 't Hooft
Institute for Theoretical Physics
University of Utrecht, Netherlands
1. INTRODUCTION
"Quantum Chromodynamics" is a pure gauge theory of fermions

and vector bosons that is assumed to describe the observed strong
interactions. To get an accurate theory it is mandatory to go
beyond the usual perturbation of expansion. Not only must we ex-
plore the mathematics of solving the field equations non-perturba-
tively; it is more important and more urgent first to find a
decent formualtion of these equations themselves, in such a way
that it can be shown that the solution is uniquely determined by
these equations. We understand how to renormalize the theory to
any finite order in the perturbation expansion, but it is ex-
1-5 )
pee ted that this expansion will diverge badly, for any value
of the coupling constant. Thus the expansion itself does not yet
define the theory. But the renormalization procedure is not known
to work beyond the perturbation expansion. A clear example of the
possible consequences of such an unsatisfactory situation was the
.. d . 6-10) .
recent supr~s~ng emonstrat~on that all non-Abel~an gauge theo-
ries have parameters e, in the form of an angle, that describe cer-
tain symmetry breaking phenomena in the theory, but never show up
within the usual perturbation expansion because they occur in the
943
944 G. 't HOOFT
combination
where g is the gauge coupling constant. This discovery marks an

increase in our understanding of non-perturbative field theory but
this understanding is not yet complete and, in principle, more of
such surprises could await us.
The situation can be compared with the infinity problems in

the older theories of weak interactions. Those problems were
solved by the gauge theories for which an acceptable and unique
regularization and renormalization scheme was found. For our pre-
sent strong interaction theory, again a "regularization scheme"
must be found, this time for "regularizing" the infinities encoun-
tered in summing the perturbation expansion.
An interesting attempt to give a non-perturbative formulation

of the (renormalized) theory is the introduction of a space-time
11-13)
lattice in var10US ways . But, here also, a proof of unique-
ness could not be given (will the continuum-limit yield one and
only one theory?) and of course the 8 phenomenon mentioned before
was not observed in the lattice scheme. So, the lattice theories
are still a long way off from answering our fundamental questions.
It is more important for us to make use as much as possible

of the important pieces of information contained in the coupling
constant expansion. Because of asymptotic freedom 14 - 16 ) this ex-
pansion tells us precisely what happens at asymptotically large ex-
ternal momenta and it would be a waste to throw that information
away.
Which tools could we use to extend our definitions? One is a

study of the theory at complex values of the coupling constant.
That this is possible in some particular cases is explained in
Section 4. We may find that Green's functions must become singular
at certain points in the complex g2 plane and stay regular at others.
CAN WE MAKE SENSE OUT OF "QUANTUM CHROMODYNAMICS"? 945
Suppose we would discover that there are only singularities on the

real axis for -00 ~ g2 < -a (unfortunately the true situation is
much less favourable). Then we could make a mapping:
~2.-+ CL =~+ 2fl o. [~ - V~+a.-i92.]

with the inverse
The whole complex plane is now mapped onto the interior of the
unit circle, with the singularities at the edge. If we rewrite
the perturbation expansion in terms of u instead of g then we
have convergence everywhere inside the circle*), which implies con-
vergence for all g2 not too close to the negative real axis. A
definite improvement. Unfortunately, the structure for complex g2
we find in Section 4 is too complicated for this method to work:
the origin turns out to be an essential singularity. Still, I
shall show how this knowledge of the complex structure may be used
to give a very slight improvement in the perturbation expansion
(Section 9).
The other tool to be used is the Borel resummation procedure,

to be explained in Section 5. A new set of Green's functions" is
considered whose perturbation expansion terms are defined to be the
previous ones divided by (n-l)! where n is the order of the expan-
sion terms. The singularities of these new functions can be found
and the analytic continuation procedure as sketched above can be
applied to obtain better convergence. There seem to emerge two
types of singularities: one due to the instantons in the theory,
the other due to renormalization phenomena. The latter are slight-
ly controversial, they are the only ones that should occur in
*) We make use of a well-known theorem for analytic functions that

says that the rate of convergence of an expansion around the
origin is dictated by the singularity closest to the origin.
946 G. 't HOOFT
quantum electrodynamics, giving that theory of n! type divergence.

The first few terms for the electron g-2 do not seem to diverge
the way suggested by this singularity. Formally we can improve the
perturbation expansion for QED but in practice the "improvement"
seems to be bad.
Our work is not finished. What remains to be done is to prove

that all singularities in the Borel variable have been found and
to find a prescription how to deal with those singularities that
are on the positive real axis. Finally, it must be shown that the
integrals that link the Borel Green's functions with the original
Green's functions make sense and a good theory is found (see Sec-
tion 9).
The problem of quark confinement can probably be related to

certain singularities on the positive real axis, because these
singularities arise from infra-red divergences (Section 8).
2. DEFINITION OF THE COUPLING CONSTANT AND

HASS PARAMETERS IN TE:'LMS OF L~~GE HOMENTUM LIMITS
This section contains the mathematical definitions of the pa-

rameters in the theory, so that the statements in the other sec-
tions can be made rigorous and free of unnecessary assumptions. It
could be skipped at first reading.
The dimensional renormalization scheme is a convenient way of

defining a perturbation series of off-mass shell Green's functions
with some coupling constant gD(W) as an expansion parameter. The
17-20)
subscript D stands for dimensionally renormalized . Each
term of the Derturbation series is finite. There is an arbitrari-
ness in the choice of the subtraction Doint W (which has the di-
mension of a mass). The theory is invariant under a simultaneous
change in gD and W provided that
(2.1)
Here we show explicitly the minus sign for the first coefficient.
This minus sign is a unique property of non-Abelian gauge fields
and is responsible for "asymptotic freedom" (at increasing ].l we
get decreasing g2, see Refs. 14-16).
21-23)
The coefficients 81 and 82 are known . Since we shall
try to go beyond perturbation expansion we must be aware of two
facts: First, the perturbation expansion is expected to diverge
for all g2 and is, therefore, at this stage, meaningless as soon
as we substitute some finite value for g2. Second, the dimensional
procedure has only been defined in terms of the perturbation ex-
pansion (Feynman diagrams). Consequently, g~ may not have any
meaning at all as a finite number. The correct interpretation of
these series is that they are asymptotic series valid for infini-
tesimal g only, or, equivalently, valid only for asymptotically
large moment: p2 O(].l2) + 00. Thus, g~ may not be so good to use
as a variable for a study of analytic structures at finite complex
values.
We shall now introduce another parameter g~, that may just as

well be used instead of g~. It is defined by the following require-
ments:
When ].l + 00, then
(2.2a)
(2.2b)
948 G. 't HOOFT
The series in (2.2b) must stop after the second term. In pertur-
bation theory these requirements have a unique solution for g~.
For instance, we get that the rest term in (2.2a) is
(2.3)
In the dimensional renormalization scheme also the mass parameter

was cut-off-dependent (of course, one cannot define such a thing
as a ~hysical quark mass parameter, which would have been cut-off
independent):
Again we define ~(~) by
(2. Sa)
and
(2.Sb)
where the series in (2.Sb) stops ~fter the al term.

21-23)
In QeD the parameters aI, 61, 62 are known
~~ = (8~2.f" (~~ -2N~ 13)

fo2. = (81t2.t2(~9 NF'3-5'1) (2.6)
.. = _C2TC2.)-i
We emphasize that the new parameters gR and ~ are better than the
previous ones, because for any theory for which the perturbation
expansion is indeed an asymptotic expansion, they are completely
finite and non-trivial. Of course, they still depend on the sub-
traction point~. At infinite ~ they coincide with other defini-
tions; at finite ~ they are finite because we can solve Eqs.
(2.2b) and (2.Sb):
(2.7a)
(2.7b)
Here ~o and mo are integration constants. They are invariant

under the renormalization group. Thus, ~o 1S a true parameter that
fixes the gluon couplings, and has the dimensions of a mass. For
each quark in the system we have a mass parameter mo. As must be
clear from the derivations, ~o and mo actually tell us how the
theory behaves at asymptotically large energies and momenta.
The value of ~o for QeD is presumably of the order of the

mass, and mo will be a few HeV for the up and down quarks, 100 HeV
or so for the strange quark, etc. For simplicity, we will often
drop the terms containing 62, and the quark masses will be put
equal to zero.
3. THE RENORHALIZATION GROUP EQUATION
Now let us consider the ~reen's functions of the theory. For

definiteness, take only the two-point functions (dimensionally re-
normal i zed)
950 G. 't HOOFT
(3.1)
. f y a renorma1
They sat1s
1zat1on . 24) :
group equat10n
[~+ (!>R(ci\;)~t + ~(9~)] Gob ( K2.d " ~~) =0 (3.2)
Here SR is the truncated, finite S function for the constant g~

as it occurs in Eq. (2.2b), but y(g~) is still an infinite series.
We wish to do something about that also. Let us first make clear
how to interpret Eq. (3.2). Consider the ~ versus g2 plane. Sup-
pose we choose a special curve in that plane, where g~ depends on
~ such that
(3.3)
then it follows from (3.2) that
or
(3.5)
That implies that, if we stay on one of the curves (3.3), then

(3.2) reduces to (3.5) which can easily be integrated. But the
integration constant will still depend on k 2 and on the curve
chosen, that is, on the constant ~o, which we get in solving (3.3),
see Eq. (2.7a). Thus we get
(3.6)
with
(3.7)
and for dimensional reasons, G(k2,~o) can only depend on the ratio
k2/~~.
For our purposes it is now important to observe the ~ollowing.
The coefficients Zo and ZI are clearly very important as g2 + 0,

but the terms z2g~ + in (3.7) can simply be absorbed in a re-
definition of the coefficients aI, . in (3.1). That way we get
new, improved functions GR that can be written as
(;.RCK1., ~/9~) = a.o(K2)+9~Q.~(Kl, ~.)+ .

=C~~ fit., t.)tPC- ~o ) G-( K1./tt!)
(3.8)
~R
In a typical example where we study the time ordered product of two

operators
'Flo) '1'(0)
corresponding to the a channel, we have

952 G. 't HOOFT
r=2 - n: -2. 9~'" ...

zo= 21 fb~ (3.9)
z.f = -"It- 2. ~; -t + 2.,.,1. ~~ 2.
4. ANALYTIC STRUCTURE FOR COMPLEX g2
In Eq. (3.8) we can write (neglecting for simplicity the S2

terms)
(4.1)
This is a function of one single parameter
x = i~2. + i.2. f'~.. ~

--;}
't(2/r-l. (4.2)
Complex x corresponds to either complex k 2 , real g2 or complex g2,

real k 2 Now, on physical grounds, we know what we should expect
at real g2, complex k 2 (Fig. 1). The singularities are at k 2 real
and negative (i.e. Minkowskian). That is when x = real + (Sl/2)-
(2n + l)ni, n integer. Choosing now k 2 real and positive we find
the same singularities at
(4.3)
They are sketched in Fig. 2.

-/. ..:-- o
-I
Fig. 1 Expected analytic structure of GR for complex k 2

The wavy line is a cut (in Baryonic channels this
cut starts away from the origin, in mesonic chan-
nels, since we have put mf = 0, the cut starts at
zero because the pion is massless). The dotted
crosses are singularities to be expected in the
second Riemann sheet (resonances).
954 G. 't HOOFT
Fig. 2 Resulting analytic structure for complex g~. The

single cut of Fig. 1 now reproduces many tImes on
semi circles. These semi-circles are only slightly
distorted due to the 62 term in Eq. (2.7a). The
arrow shows the region where perturbation expansion
is done. The cut on the left is due to the Z-fac-
tor in (3.8).
The conclusion of this section is obvious: we find such a bad

accumulation of singularities at the origin that the analytic con-
tinuation procedure given in the introduction will never work. We
must look for a more powerful technique.
5. BOREL RESUMMATION
We now assume that our Green's functions can be written as a

Laplace transform of a special type
(5.1)
F(z) can be found perturbatively: If
(5.2)
then
(5.3)
where the 0 function is understood to be included in the integral

(5.1). One can check this trivially by inspection. The importance
of this is that the series (5.3) converges much faster than (5.2).
Contrary to (5.2) it may very well have a finite radius of conver-
gence (this is at present believed to be the case for all renor-
malizable field theories). If F(z) can now be analytically con-
tinued to all real positive z, and if, for some g2, the integral
(5.1) converges, then the series (5.2) is called Borel summable.
e will call F(z) the Borel function corresponding to the Green's
function GR(g2).
First, we wish to find out where we can expect singularities

in F(z). Let us illustrate an interesting feature in the case of
an over-simplistic "field theory", namely field theory at one space-
time point. Remember that field theoretical amplitudes can be writ-
ten as functional integrals, with a certain number of integration
. bl es at eac h space-t1me
var1a . . 10) If we have one space-t1me
p01nt
point and one field, then there is just one integration to do

956 G. 't HOOFT
(5.4)
and the factor g in front is just

for convenience. g2 comes out this way as the usual perturbation
parameter. Note the minus signs in the integrand. This anticipates
that we shall always consider Field theory in Euclidean space-time.
Let us rescale the fields, and the action,
=_is'(A') (5.5)
~1
g'(A'): iCA,)L+V(A')
2.
Our integral becomes
(5.6)
Comparing this with the Borel expression (5.1), we immediately find

F (z) :
F(z) : JriA' ~ (z _ S'CA') (5.7)
Thus at given z we must find all solutions of S'CA) z, which we

can call A.(z). The result of the integral is
1.
(5.8)
This outcome reveals the singularities in the z plane: we must

find the solutions of
'JS'(A) = 0 (5.9)
7JA
to be recognized as the classical field equation of this system.
At a solution Aof (5.9) we have
%=S'(A)
[ C>S'] -1 [2 ';)2.SI ex _%)]-of/2- (5.10)
"i>A ~ C1A2..
Q.s z-+ x
and so we find a square root branch point at z = z.
Consider now multidimensional integrals of the same type. The

integral (5.7) then corresponds to an integral over the contours in
+ +
A space defined by the equation S'(A) = z. You get singularities
only at those values of z that are equal to the total action S' of
a solution of the equation
;) S'
~=O (5.11)
'JA
because those are the contours that shrink to a point (in the case
of a local extremum) or have crossing points (saddle points).
Again, Eq. (5.11) ~s nothing but the classical Lagrange equation
for the fields A. Conclusion: to find singularities in F we have
to search for finite solutions of the classical field equations in
Euclidean space-time. Their rescaled action S' corresponds to
singularity points ~ in the z place for the function F. In general,
these singularities are branch points. In our actual four-dimensional
958 G. 't HOOFT
field theories that are supposed to describe strong (or weak and
electromagnetic) interactions; such solutions indeed occur, and are
called "instantons", because they are more or less instantaneous
and local in the Euclidean sense 25 ,26,6,lO). Their action (in the
case of QeD) is S' = 8n 2n, where n counts the "winding number"a-lo)
and so we may expect singularities in the complex z plane at z =
= 8n 2 n.
6. UNIVERSALITY OF THE BOREL SINGULARITIES
The student might wonder whether the conclusions of the pre-

vious sections were not jumped to a little too easily. The con-
nected Green's functions in field theories are not just multi(in-
finite) dimensional integrals but rather the ratio of such inte-
grals with some source insertion and an integral for the vacuum,
and then often differentiated with respect to those source inser-
tions. Do all these additional manipulations not alter or replace
these singularities and/or create new ones? Do different Green's
functions perhaps not have their own singular points?
Let us for a moment forget the renormalization infinities, to

which we devote a special section. Then the answer to these equa-
tions ~s reassuring. ~1ultiplications, divisions, exponentiations
can be carried out, after which we shall always find the singulari-
ties back in the same place as they were before, possibly with a
different power behaviour. To understand this general property of
Borel transforms, let us formualte some simple properties.
Let
(6.1)
Then, if
then
(6.2)
And let
then
(6.3)
Here the symbols are there just to tell us to l~ave the 0 symbols
out of the integration. It is easy to show that if (5.3) is solved
iteratively, then the series converges for all z, as long as Fl
stays finite between and z.
Now note that we may choose the contours (O,z) so that they
avoid singularities. Only if z is a singularity of either Fl or
F2, or both, then F 3(z) in Eq. (6.2) will be singular. Also F2(Z)
in Eq. (6.3) is only singular if F (z) is singular. Note, however,
that if a singularity lies between 0 and z then the contour can be
chosen in two (or more) ways, and, in general, we expect the out-
come to depend on that. Thus if we start with pure pole singulari-
ties, they will propagate as branch points in the other Borel func-
tions.
In quantum field theories, the Green's functions are related

through many Schwinger-Dyson equations and Ward-Slavnov-Taylor
identities. Since singularities survive the multiplications and
divisions in these equations without displacement, they must occur
in all Borel-Green's functions at the same universal values of z
960 G. 't HOOFT
(unless miraculous cancellations occur; I think one can safely ex-

clude that possibility). These singularities will, in general, be
of the branch-point type.
In particular, those singularities that we obtained through

the solutions of the classical equations, will stay at the same
position for all Green's functions.
7. SINGULARITIES IN F(z) DUE TO INSTANTONS
Let us consider these classical equations ~n Euclidean space-

time for the various field theories. First take A4 theory, for
simplicity, without mass term. Rescaling the fields and action the
usual way
(7.1)
we have
(7.2)
I t turns out 25) t h at a pure1

y ~mag~nary so I
ut~on
.
ex~sts
f or t h e
equations ~ S/rl = 0 (here, tlindicates derivative in the Euler

sense), namely
(7.3)
Here p is an arbitrary scale parameter (after all, our classical ac-

tion ~s scale invariant).
In spite of this solution being purely imaginary, it is impor-

tant to us because it indicates a singularity in F(z) away from the
positive real axis. The corresponding value for S' is
(7.4)
So the singularity occurs at z = -l6n 2 , indeed away from the posi-

tive real ax~s. Such singularities are relatively harmless, since
F is only needed for positive z. We may invoke the analytic con-
tinuation procedure sketched in the Introduction to improve conver-
gence for the series in z.
Now, let us turn our attention to Quantum Chromodynamics. Here

we have a real solution in Euclidean space:
2. 't1 )( \)
Au = ~ A
0.' Q.
='OCf(~ (7.5)
,- f4- )( 2. + e,1
. 6,27)
where n are certa~n real coeff~c~ents and p ~s aga~n a free
a~JV
scale parameter. One finds for the action
(7.6)
Thus z = 8n 2 is a singularity on the positive real axis. In fact,

we can also have n instantons far apart from each other, so we also
expect singularities*) at z = 8 n. Now, Green's functions are ob-
tained from F(z) by integrating from zero to infinity, over the
positive real axis. Do the singularities on the real axis give un-
surmountable problems? I think not, although the correct prescrip-
tion will be complicated. A clue is the following. The single-
instanton contribution to the amplitudes has been computed directly
*) A more precise analysis suggests that only those multi-instantons

with zero total winding number (that is, as many instantons as
anti-instantons) will give rise to ordinary singularities that
limit the radius of convergence of F(z). The others give dis-
continuities rather than singularities.
962 G. 't HOOFT
in the small coupling constant limit. A typical result goes

like 27 - 30 )
(7.7)
That is already a Green's function, the one we would like to ob-

tain after integrating
(7.8)
A function F(z) that yields (7.7) exists
(7.9)
Indeed, a "singularity" at z = 8n 2 We see that, since all Green's

functions will show the same exponential in their g dependence,
the universality theorem of the previous section is obeyed. What
is important is that by first computing (7.7) one can short-circuit
the problem of defining an integration over such singular points.
Thus the instanton singularities at the right-hand side on the
real axis will not destroy our hopes of obtaining a convergent
theory. The reason is that the physics of the instanton is under-
stood. The situaiton is less clear for the other type of singulari-
ties that we discuss in the next section.
8. OTHER SINGULARITIES IN F
In principle, the instanton-singularities in F can also be un-

derstood within the context of ordinary perturbation expansion, by
31)
a stat1st1cal treatment of Feynman d1agrams We do not show the
derivation here, but the following argument has been given. In the
previous section, we have never bothered about the renormalization
procedure that is supposed to make all diagrams finite. Suppose we
had a strictly finite theory, with bounded propagators, bounded in-

tegrals and all that. Individual diagrams in such a theory are then
bounded by a pure power law as a function of their order n. The
only way that factors n! can arise is because there are n! diagrams
at nth order and they may not cancel each other very well. This is
how in the statistical treatment the instanton singularity occurs.
But in realistic four-dimensional renormalizable field theories, the
power law for individual Feynman diagrams no longer holds. A simple
example is quantum-electrodynamics. We consider the diagrams of the
type shown in Fig. 3.
Fig. 3 Fourth member of a subclass of dia-

grams discussed in this section.
It is the class of diagrams with n electron bubbles in a row, which

in itself closes again a loop. It is well known that each electron
bubble separately behaves for large k 2 as
(8.1)
and each propagator as (k 2 )-1. Thus, for large k 2 the integrand in

the k variable behaves as
964 G. 't HOOFT
(8.2)
where a is some fixed power. After having made the necessary sub-
tractions to make the integral converge, and in order to obtain
physically relevant quantities, such as a magnetic moment, the
leading coefficient a becomes 3 or larger. Let us replace log k 2
by a new variable x, then (8.2) becomes proportional to
(8.3)
Thus the integral over x will grow as n 7 00 like
(8.4)
A more precise analysis shows that C should be proportional to the

first non-trivial 6 coefficient
(8.5)
In the expansion for F(z) the factor n! is removed, as usual. It

is clear that a new singularity develops at
(8.6)
It seems to be a universal phenomenon for all field theories, and

not related to any instanton solution. Our definition for 61 was
positive for asymptotically free theories and negative otherwise.
So, the singularity is at negative real z and therefore harmless if
our theory is asymptotically free, but for non-asymptotically free
theories such as QED and A4, we have singularities on the positive
real axis. Since a detailed understanding of the ultraviolet
behaviour of non-asymptotically free theories is lacking, there may

exist no cure for these singularities then. This is in contrast
with the instanton singularities.
.
An ~mportant .
observat~on
h as b een made by G. ..
Par~s~
32)
The
ultraviolet behaviour of A~4 and QED are well understood in the
limit N + 00, where N is the number of field components. A syste-
matic study of the singular point (8.6) is then possible. Parisi
found in A~4 theory a conspiracy between diagrams such that the
first singularity at a 3 cancels. In total the integrals do be-
have as (8.2) but with a > 3, after all necessary subtractions. At
present, it is not understood whether this conspiracy is accidental
for A~4 theory with N components, or whether it is a more general
phenomena. It does seem that only the first singularity may be
subject to such cancellations.
In Figs. 4 and 5 we show the complex planes for the Borel

variables z in A~4 theory and in QED, respectively. The singulari-
ties discussed in this section are called "renormalons" for short.
?
-6 -3 4 6 8
instantons renormalons
Fig. 4 Singularities in the Borel z variable for A4.

The units are 16n 2 /3. The question mark denotes
the singularity that may be cancelled according
to Parisi's mechanism.
The situation for QeD is more complex. Not only do we have the
renormalons at points on the negative real axis but also there are
such singularities on the positive real axis. They are due to the
infra-red divergence of the theory. The mechanism is otherwise the
966 G. 't HOOFT
?
1 2 3 4
renormalons
Fig. 5 Singularities for QED. Here the units are 3TI,

if a is the original expansion parameter.
IR divergencies
?
)) instantons
+
renormalons -4
lr,
Fig. 6 Borel z plane for QeD. The circles denote IR

divergences that might vanish or become unim-
portant in colour-free channels.
the same as discussed for the unltraviolet singularities. The in-

stanton singularities corne on top of these (Fig. 6).
An interesting speculation is that these infra-red singulari-

ties are only surmountable in colourless channels, but the integra-
tion over these singularities becomes impossible in single quark-
or gluon- channels. It is likely that these singularities are re-
lated to the quark confinement mechanism.
9. SECOND BOREL PROCEDURE
Many features of the singularities in the complex z plane of

the Borel functions F(z) are still uncertain and ill-understood.
But from the foregoing we derive some hopes that it will be possible
to obtain F(z) for 0 ~ z < 00 for asymptotically free theories, such
as QCD. The only thing to be investigated then is how the integral
in
(9.1)
behaves at oo Does the integral converge? The answer to this is

almost certainly: no. Consider massless QCD and its singularities
in complex g~ plane as derived in Section 3. According to Eq. (3.3)
there are singularities when
where the real number may be arbitrarily large. Substituting that

in Eq. (9.1) we find that
J F(~) e _x. (uai + ~.Pti

00
2.
(21Jl+ ~
d.x. (9.3)
o
must diverge. We had assumed that the singularities at finite z did

not give rise to divergences. So F(z) must diverge at large z worse
than any exponential of z. Note that (9.3) contains an oscillating
term. It is likely then, that at z + 00, F(z) does not only grow
very fast, but also oscillates with periods 4/S1 or fractions there-
of.
Can we cure this disease? We have no further clue at hand which

could provide us with any limit on the large -z behaviour of F. But
there is a way to express the unknown Green's functions in terms of
968 G. 't HOOFT
a more convergent integral than (9.3). Let us treat the divergent

integral (9.1) on the same footing as the divergent perturbation ex-
pansions which we had before. We consider a new, better converging
integral
A x/2
WCs) = LCO F(x.) dl.. S'-'" (9.4)
We may hope that this has a finite region of convergence, from which
we can analytically continue. Note the analogy between (9.1) and
(9.4) on the one hand, and (5.2) and (5.3) on the other. The inte-
gral relation between Wand G, analogous to (5.1), is
(9.5)
Now, remembering that instead of varying g2 we could vary k 2 , re-

placing
(9.6)
So that, ignoring the Z factor that distinguishes G from GR

(see Eq. (3.8)), one gets
(9.7)
Now we can easily prove that, if our theory makes any sense at all,
there may be no singularities in W(s) on the positive real axis, ane
the integral (9.7) must converge rapidly. Thus, if the integral
(9.4) makes sense, then our problems are solved. The proof goes as
follows:
G(k 2 ) satisfies a dispersion relation: it lS determined by its

imaginary part. We have, at k 2 = _a 2 , a > 0,
(9.8)
where p is usually a positive spectral function. Substituting (9.8)

into (9.7) we get
-
. (2.)
2. It&.f Q; =
JOO2CA) d. wW(CAJ)
2. [_two..
e -e i.c...>4l,]
o,z. 0
= _2. f oo e
_00
LWA
W.
CO W(w
~
..) ~
ClW
(9.9)
Thus, pea) and W(w 2 ) are each other's Fourier transform. The in-
verse of (9.9) lS
w(s) (9.10)
A possible singularity at a 2 = 0 is an artefact of our simplifica-

tions and can be removed. It is important to observe that (9.10)
severely limits the growth of W(s) at large s so that (9.7) is
always convergent.
Conclusion: this section results in an improvement on per-

turbation theory. The physically relevant quantities can be ex-
pressed in terms of integrals of the type (9.4), which converge
better than the original ones of type (9.1). It lS not known whether
this improvement is sufficient, i.e., whether (9.4) actually conver-
ges in some neighbourhood of the origin.
Even if the important open questions mentioned in these lec-

tures cannot be answered we think that refinememnt of these tech-
niques will lead to an improved treatment of strong coupling
theories.
970 G. 't HOOFT
REFERENCES
1) F.J. Dyson, Phys. Rev. 85 (1952) 861.
2) L.N. Lipativ, Leningrad Nucl. Phys. lnst. report (1976)

(unpublished) .
3) E. Brezin, J.C. Le Guillou and J. Zinn-Justin, Phys. Rev. D15

(1977) 1544 and (1977) 1558.
4) G. Parisi, Phys. Letters 66B (1977) 167.
5) C. ltzykson, G. Prisi and J.B. Zuber, Asymptotic estimates ~n

Quantum Electrodynamics, CEN Saclay preprint.
The singularities I discuss in Section 8 of my lectures were
assumed to be absent in this paper.
6) G. 't Hooft, Phys. Rev. Letters 37 (1976) 8.
7) A.}1. Polyakov, Phys. Letters 59B (1975) 82 and unpublished

work.
8) C. Callan, R. Dashen and D. Gross, Phys. Letters 63B (1976) 334.
9) R. Jackiw and C. Rebbi, Phys. Rev. Letters 37 (1976) 172.
10) See S. Coleman's lectures at this School.
11) K.G. Wilson, Phys. Rev. D10 (1974) 2445.
12) J. Kogut and L. Susskind, Phys. Rev. Dll, 395 (1975).

L. Susskind, lectures at the Bonn Summer School (1974).
13) S.D. Drell, M. Weinstein, S. Yankielowicz, Phys. Rev. D14

(1976) 487 and DL4 (1976) 1627.
14) G. 't Hooft, Marseille Conf. on Renormalization of Yang-Mills

fields and applications to particle physics, June (1972)
(unpublished) .
15) H.D. Politzer, Phys. Rev. Letters 30 (1973) 1346.
16) D.J. Gross and F. Wilczek, Phys. Rev. Letters 30 (1973) 1343.
17) G. 't Hooft and M. Veltman, Nuclear Phys. B44 (1972) 189.
18) C.G. Bollini and J.J. Giambiagi, Phys. Letters 40B (1972) 566.
19) J.F. Ashmore, Lettere a1 Nuovo Cimento 4 (1972) 289.

20) G. It Hooft, Nuclear Phys. B6l (1973) 455.
21) D.R.T. Jones, Nuclear Phys. B75 (1974) 531.
22) A.A. Belavin and A.A. Migdal, Gorky State University pre-
preprint (January 1974).
23) W.E. Caswell, Phys. Rev. Letters 33, (1974) 244.
24) S. Coleman, lectures given at the "Ettore Majorana" Int.

School of Subnuclear Physics, Erice, Sicily (1971).
Note: we drop the inhomogeneous parts of the renormalization
group equation, which can be avoided according to later
formulations on the renormalization group (Ref. 20).
25) S. Fubini, Nuovo Cimento 34A (1976) 521.
26) A.A. Belavin et al., Phys. Letters 59B (1975) 85.
27) G. It Hooft, Phys. Rev. D14 (1976) 3432.
28) F.R. Ore, "How to compute determinants compactly", MIT pre-

print (July 1977).
29) A.A. Belavin and A.M. Polyakov, Nordita preprint 77/1.
30) A.M. Polyakov, Nordita preprint 76/33 (Nuclear Phys: in press).
31) C.M. Bender and T.T. Wu, Phys. Rev. Letters 27 (1971) 461;
Phys. Rev. D7 (1972) 1620.
32) G. Parisi, private communication.

DIS C U S S ION
CHAIRMAN: Prof. G. 't Hooft
Scientific Secretaries: E. Mottola, C. Tao
DISCUSSION No.1
BAULIEU:
How do you obtain the circles in the complex g2

plane?
't Hooft:
Originally, in the theory, k 2 and p2 are the only

free parameters. The coupling constant g2 is related to
~2 through the renormalization group equation:
(higher order terms which

are not important here)
the solution of which is:
9:(1A) = ~1 4?r ( ~o)

where ~o is a free integration constant with dimension
of a mass. Now the point is that I am free to choose the
~ anyway I like so that I can just as well consider ~
fixed and g(~) the free parameter or the integration
constant ~o as the free parameter.
Simple dimensional arguments show that any Green's
funct ion can depend only on \("Z / \l'Zo=}("
apart from some
renormalization factors in front, which I'll leave out
for simplicity.
973
974 DISCUSSION
Also from elementary physi~s, we know what the anal-

ytic structure in the complex k plane should be: just
poles and cuts on the negative real axis. So for ~ =-a,
one such negative real singular point, we may as well
consider k 2 real and
parameter x= JC.2/p!
\1:
complex since there is only one
What does complex \A: imply about
g2?
If \l!is a real negative number, then to\ r~:=. l2nt-I)17t+
+(real)so that in the 1/~~ plane, the singular~ties must
lie on lines.
t.Y~~
-----+---- ~ilt
1.'7t
----~-------3i~
Now, if we go to the g2 plane, the result is just a set

of circles by inversion about the origin.
BUDNY:
You mentioned that instantons have the effect of

intrOducing the physical parameter a in QCD. Since we
cannot get this from perturbation theory, there is the
possibility of many such parameters. This would have the
effect of preventing us from calculating things just as
in non-renormalizable theories. Is there any reason to
hope that there is a small number of such non-perturbative
physical parameters?
't HOOFT:
It is possible that there is an infinity of those

parameters.
COLEMAN:
There are exact soluble models having such parameters
and, in those cases, there is typically only one. There
may, of course, be any number but there is no reason
whatever to expect that there are more.
DISCUSSION 975
WIGHTMAN:
Please describe the relation between your analysis

of the singularities in the complex g2 plane and
N. Khuri's study of the trajectories in the complex g2
plane.
' t HOOFT:
I am afraid I am not familiar with that and I can-

not answer.
COLEMAN:
Khuri did not use Gerard's clever trick of defining

the coupling constant in such a way that the Callan-
Symanzik functions were exactly known. Therefore his
analysis is bogged down quite a lot in technical assumpt-
ions about the analytic behaviour of the Callan-Symanzik
functions. Gerard essentially shortcircuits about 95%
of Khuri' s work.
PAFFUTI:
Is the Borel summation the "unique" way of regular-

ization of the perturbation expansion?
't HOOFT:
This is exactly the question I am trying to investi-

gate in these lectures. There may be two sources of
trouble that make the method ambiguous, not unique. One
is that while integrating over z, from 0 to 00 , one might
encounter singularities on the positive real axis. And
our problem is then to find a unique prescription of
integrating over them. The second is that the integral
might well diverge at 00 . Any ad hoc cut-off there,
would of course make the method ambiguous. I will elab-
orate on this tomorrow.
PREPARATA:
Could you please clarify the differences between QCD

and QED, as far as the coupling constant singularities
are concerned?
976 DISCUSSION
It seems to me that the difficulties you mention

come from imposing that the theory exhibits a spectrum
of the dual model type.
COLEMAN:
You don't need to make any assumption about rising

Regge trajectories or any confinement to get that circle
of singularities. It suffices to say that you have nor-
mal thresholds on the real axis. For sufficiently large
timelike k 2 or cuts on the left hand side, you certainly
have a pion and that certainly means that you get 2~
thresholds, 41t thresholds, 61t thresholds and ad infini-
tum. Those are branchpoints and not poles but that's
irrelevant. When Gerard makes his transformations, they
go on that circle tangent to the real axis and converge
on the origin. So because of multiparticle states,
resonances or not, you still have zero opening angle.
A cut is movable but the branchpoint is immovable,
so you always have it. These tricks are also based on
the fact we have an asympotically free theory with a
negative leading term. In QED, I don't know, but maybe
Gerard can speak for that.
't HOOFT:
In QED, the same analysis would give circles that

go the other way round, so that you would have an open-
ing angle to the right.
However, there are difficulties in the procedure
here due to non-asymptotic freedom and the presence of
a mass term.
TOWNSEND:
If the subtraction point can be chosen, so that

@>=-~I '34 -t f2~E. exactly, and if ~1 and ~2 are sub-
traction independent, why is it that the whole theory,
and zeroes of ~ in particular, cannot be determined
from perturbation theory by calculation of ~ and ~2 ?
't HOOFT:
What is wrong with your suggestion is that this

mathematical trick only gives the positions of the sing-
DISCUSSION 977
ularities, which is only a very small part of all the

information that you want. For instance, the residues
of the poles are not given by those two coefficients.
WEEKS:
Could you please state a working version of the

Euclidicity postulate?
't HOOFT:
A vacuum expectation value of two operators at

different times in Euclidian space is:
- tt
<All:) I A"Co,) :: L <A (0) lEi? <E\ \ A/Co') e I
E
and in Minkowski space: ~ . t=.t
- I L..j
<Alt-')IA'(oJ) = L (AlC) IEi/<Ei \A'(t))le .
'E;
One should be the analytic continuation of the other.
COLEMAN:
The statement that you analytically continue any

time ordered product into Euclidian space, i.e., imagin-
ary time, is just to say that all the energies are posit-
ive : because that means that all your sums get more and
more convergent as an oscillating exponential becomes a
damped exponential.
There is a precise mathematical theory, the
Osterwalder-Schrader theory, that tells you, if you have
a set of Green's functions in Euclidian space, what ass-
umptions they must obey in order to be continued analytic-
ally back to Minkowski space, such that this analytical
continuation obeys all Wightman's axioms, including
positivity.
Now, what happens when you know the Euclidian Green's

functions only approximately?
There, it depends on what you are computing. If
you are computing energies, masses, or residues of poles,
or matters of that kind, then you can show that the an-
alytical continuation is stable. You get the values
approximately, but you cannot introduce factors that are
978 DISCUSSION
small in the Euclidian region and get large when you

continue them in the Minkowski region. That can be
demonstrated explicitly.
For quantities like S-matrix, I don't know.
PAAR:
You showed that in QED an expression for

<,~_2)/2. = ~ ~ aV\ (ri y'
has about 137 decreasing terms which make the series
look convergent and then ever increasing terms which
make it diverge. Why should we believe that the inter-
mediate result after 137 terms is the correct result?
't HOOFT:
the integral 1
Compare for instance the perturbation expansion for
DO d).
e~pl-~/A) -1+'2.
o
Here, it is not difficult to show thft the true value
always lies in between the nth and t 1e(n+l)th approxi-
I
mation, for all n. So the best result is obtained if

you do the expansion until the nth correction is equal
in magnitude to the(n+l)th. There is still an inaccuracy
We assume that this strategy also makes sense in Quantum
Field Theory.
WIGHTMAN:
The usual theorems on Borel summability say that a

function f, analytic in a sectorial neighborhood of the
origin _"%-.<a.{"~'Z< 1T'~+ , c< I'ZI< R, bounded in the
closure of the vectorial neighborhood, and such that
-ftY\) (0+) exist, n=O,I,2,3, ... has a Borel summable
Taylor series. The horn-like region which you have dis-
played is not of that type. How do you make it plausible
that your series is Borel summable?
't HOOFT:
This is indeed an important problem and the answer

is not obvious, and I will lecture on this tomorrow.
DISCUSSION 979
DISCUSSION No.2 (Scientific Secretaries: B. Celmaster,

E. Mottola, C. Tao.)
CARTER:
Could you explain again where the universality of

the singularity structure of the Borel function f(z)
comes from? I understand that the instant on singular-
ities are located by stationary values of the action and
are therefore universal. In QED, I also understand that
the renormalon poles are located by ~, and hence are
also universal. Are these observations the basis for
"universality" or is the statement something very general?
't HOOfT:
There are different ways to observe this universal-

ity property. One is, as you say, to notice the fact
that the instanton action is a single universal number
and ~, is the same for all Green's functions: But there
is an independent and more general way to see this by
considering the Dyson equations among the Green's funct-
ions. We have corresponding equations between the Borel
transforms of these. You see then, that barring miracu-
lous cancellations, if any Green's function has a singu-
larity somewhere, the singularity will recur at the same
spot in all other Green's functions. I must mention that
we do not yet understand everything here. There might
be other singularities besides the instantons and the
"renormalons".
LIPKIN:
What is the basic physical difference between QED

and QCD? Is it the large coupling constant or the non-
Abelian character of QCD that is most significant? In
particular, why are quarks confined in QCD when elect-
rons are not in QED,
't HOOfT:
The basic difference between QED and QCD arises from

the non-Abelian character of QCD. However, the models
one encounters in weak interaction theory, Weinberg-Salam
for example, are also non-Abelian gauge theories. Yet
the Weinberg-Salam model is not so very different from
980 DISCUSSION
QED. The reason is that the spontaneous symmetry break-

ing affects the non-Abelian gauge field properties of
the theory in such a way that there is a closer resembl-
ence to conventional Abelian QED. SO, I would say that
QCD and QED differ because QCD is not only non-Abelian
but also an unbroken gauge field theory.
Concerning your question on confinement, opinions
vary and the answer is still unknown. My personal feel-
ing is that the essential infra-red divergence in "col-
ored" amplitudes will make the quark or gluon poles dis-
appear entirely in the completely solved theory and that
this implies confinement. Such a thing can only occur
in a theory with this particular behaviour under scale
transformations.
LIPKIN:
We know that the physics of QED is the same for

C(= 1/137 as for cl = 10- 25 The hydrogen atom would have
the same spectrum; only the scale changes. Is this also
true for QCD? Would quarks be confined in QCD even if
g2= 10- 25 or is there a phase transition?
't HOOFT:
We can perform scale transformations until the~s

effective coupling constant is 10- 25 (which is 10~o
GeV). Quark confjnement still occurs, but at that scale
it occurs at 10. 0 & units of length; at those energies,
the dynamics are very much like the behavior at 10 137 GeV.
This is just the renormalization group statement of
scaling, which is very similar to your remark on the
hydrogen atom.
LIPKIN:
How do you decide whether or not there is a phase

transition?
't HOOFT:
The effective coupling constant of the theory is a

smooth function of energy scale and all amplitudes
exhibit a dependence on g2 which shows no evidence of a
phase transition singularity, except possibly at g=O.
DISCUSSION 981
Confinement formally takes place already at g= 10- 25 so

that we do not expect any phase transition in QCD as
long as the coupling constant is non-zero.
Some people argue that QED has a phase transition
towards confinement at some large value of ~ but that
happens only if you put QED on a lattice which explicitly
introduces a length scale into the theory.
WEILER:
Are there physical interpretations for the positions

of the singularities of (f(z)Qco? In particular; why
are the instanton singularities only on the positive real
axis, and why are the singularities at multiples of gX~ ?
I would have expected the positions to depend on the
color group theoretic factors, as do the IR and UV sing-
ularities at 41."
etc.
't HOOfT:
fez) is an integral transform of a Green's function

of the theory so its singularities have an interpretation
entirely different from the singularities in the original
Green's function in the g~ or k 2 plane.
COLEMAN:
As I will show in my lecture tomorrow, 8n~ is a

group theoretic factor. However, due to Bott's theorem,
it is independent of the particular group.
't HOOfT:
Note that the relative positions of instanton and

renormalon singularities do depend on the group as well
as on the fermion representations, etc.
BOHR:
My question concerns the singUlarities of the Borel

function at infinity. Can you show more carefully how
the convergence of G(k 2 ) is obtained by introducing the
function W(s)?
982 DISCUSSION
It HOOFT:
I do not really know a priori whether or not my

definition of W(s) in terms of the z-integration conver-
ges; in any case, its convergence properties can only be
better than the integral for G(k 2 ), not worse. However,
if it does still diverge, the story stops here. On the
other hand, the relation between W(s) and G(k 2 ) or 9(m 2 )
is a completely convergent, well-defined Fourier trans-
formation. Thus, we obtain through this relation a
convergent re-definition of G(g2).
TOWNSEND:
What is the importance of the distinction between

instanton poles of F(z) which appear on the negative real
axis as opposed to the positive real axis? Is there a
connection with asymptotic freedom?
It HOOFT:
DO
We have to consider the integral J F(z) exp(-z/g2)

dz. If there are poles of F(z) on theOpositive real
axis this causes the integral to be ill-defined. This
is not the case if the poles are on the negative real
axis. Whether the instanton poles are on the positive
or negative real axis is almost certainly completely
unrelated to asymptotic freedom, because there are
theories which are not asymptotically free but which
have instanton poles on the positive real axis.
SHOULD WE BELIEVE IN QUANTUM FIELD THEORY?
A.S. Wightman
Joseph Henry Laboratories of Physics
Princeton University, Princeton, NJ 08540
The following is a straightforward attempt to answer the ques-

tion stated in the title and posed by Professor Zichichi in his
opening lecture of the School. Its perspective is to some extent
historical as is appropriate, in my opinion, for part of an attempt
to assess where we are and where we are going in the study of sub-
nuclear matter. If there are readers who are not aware of the in-
herent fallibility of such attempts to learn from history, they
should be reminded that the method involved is somewhat reminiscent
of that used by the celebrated wise men of Chelm, who, when they
wished to record the location of an especially good fishing spot,
carefully marked the place on the boat from which they were fishing.
WHAT DOES IT MEAN TO BELIEVE IN A PHYSICAL THEORY
In answering this question it is useful to consider a family

of examples: the old quantum theory, the Boltzmann equation in the
theory of dilute gases, Maxwell's (classical) theory of electricity
and magnetism, and quantum mechanics. These examples have the vir-
tue that they are fully mature theories with their main features
983
984 A. S. WIGHTMAN
understood. (That is not the case for relativistic quantum field

theory as I will discuss in detail.) There are several useful dis-
tinctions illustrated by this list.
First, unlike the others, the old quantum theory is internally

inconsistent. It combines Maxwell classical theory of electricity
and magnetism and Newtonian or relativistic mechanics with quanti-
zation conditions. There is a blatant contradiction between its
description of the emission of radiation in terms of quantum jumps
and that predicted by its equations of motion. It need not be empha-
sized to students of the parton model that an internally inconsis-
tent theory can, in spite of its inconsistency, be exceedingly use-
ful in the development of physical understanding. Certainly. the
history of quantum theory illustrates the point. The main problem
of atomic mechanics, as seen by those who were to create modern
quantum mechanics, was to obtain a theory that would give the suc-
cesses of the old quantum theory (discrete spectral lines, Balmer
formula for the spectrum of hydrogen-like atoms, etc.), without its
inconsistencies. Whatever the usefulness of internally inconsistent
theories, one hopes and expects that in the end they will be re-
placed by internally consistent theories. The rest of the list are
such examples.
The Boltzmann equation in the theory of dilute gases illustrates

a second distinction. It defines a model which is internally con-
sistent but externally inconsistent in the sense that it is incon-
sistent with the laws of statistical mechanics, except in a certain
limit in which the density of the gas goes to zero but the range of
interaction of the molecules increases as the reciprocal square root
of the density. Thus, as applied to actual gases, the Boltzmann
equation is inconsistent with the laws of statistical mechanics.
SHOULD WE BELIEVE IN QUANTUM FIELD THEORY? 985
Quantum mechanics illustrates yet a third distinction. It is

a general theory applicable in principle to any physical system in
contrast with the two preceding examples which are theories of spe-
cific systems. Quantum mechanics is internally consistent when ap-
plied to non-relativistic atoms, but the consistency has to be re-
examined anew for each area of application. For example, it is far
from clear whether there is a consistent quantum theory of gravita-
tion which is in any reasonable sense the quantization of Einstein's
theory.
A mature physical theory typically has a well-defined area of

experimental validity. For example, Maxwell's (classical) theory
of electricity and magnetism seems to describe electromagnetic phe-
nomena to the extent that quantum effects and effects of space-time
curvature are negligible. Incidentally, it is also typical of
mature physical theories that they can be put in a precise mathe-
matical form. This possibility is, in fact, what gives us abso-
lute assurance of their internal consistencyl).
Now, bearing these distinctions in mind, let us discuss the

question of belief for the listed examples of theories. Certainly,
whatever meaning is attached to the word believe it is reasonable
not to believe in internally inconsistent theories. Thus, nobody
believes in the old quantum theory. One only regards it as a use-
ful model. Similarly, it seems unlikely that anyone would want to
define believing in a physical theory so that one would believe in
the Boltzmann equation. It is just an internally consistent model
which is only approximate according to the more accurate results of
statistical mechanics. On the other hand, it seems reasonable to
define belief so that one believes in Maxwell's theory and non-
relativistic quantum mechanics.
986 A. S. WIGHTMAN
The definition that emerges from this inductive discussion is

this: one believes in a mature physical theory if
a) it is internally consistent
b) there is a significant range of experiments which it
describes accurately.
This definition is evidently somewhat arbitrarYt but it has the ad-

vantage of being objective. In any case t it provides the criterion
used in the following discussion.
For theories in the process of construction or investigation t

or whose consequences are still being tested experimentally, the
question whether one should believe in the theory is one of betting
odds. How likely is it that, after the theory has become mature,
it is one in which one must believe according to the above defini-
tion? Evidently, the probability arguments that one can bring to
bear on the matter may be either theoretical or experimental or
both.
With these definitions in hand, I now turn to quantum field

theory.
A LITTLE OF THE EARLY HISTORY OF QUANTUM FIELD THEORY
The first quantum field theory constructed was that of a free

field in two-dimensional space-time (quantization of the vibrating
string) worked out by Born, Heisenberg t and Jordan in the second
.
paper ever wr~tten on quantum mech an~cs
. 2) The auth ors note d t h at
the infinite number of degrees of freedom did not give rise to any
special difficulty; each normal mode of vibration of the string
had only to be quantized as a harmonic oscillator. Next Dirac quan-
tized the electromagnetic field in the Coulomb gauge and coupled it
to Schrodinger electrons 3 ). FinallYt the equations of relativistic
quantum electrodynamics were written down by Heisenberg and Pauli q ) .
It is a striking fact that in the sense of the above defini-

tion the creators of quantum electrodynamics did not believe in it
because they thought it was internally inconsistent. They knew that
the classical electrodynamics of point electrons is afflicted with
difficulties arising from the infinite electromagnetic self-energy
of point electrons. Since quantum electrodynamics is just a quan-
tized version of such a classical theory, they supposed that quan-
tum electrodynamics would contain the same divergent self-energies.
The first calculations of the electron self-energy in perturbation
theory supported this view; they yielded a divergent electron self-
energy. However, to the suprise of all, it turned out that the
self-energy depends only logarithmically on the ultra-violet cut-
off, instead of giving a power as in classical theory. There were
other pleasant surprises. Processes calculated in Born approxima-
tion yielded excellent agreement with experiment (Compton effect,
pair production, bremsstrahlung, the absorption and emission of ra-
diation, etc.). An unpretentious theory originally set up to pro-
vide a quantitative account of the effects of retardation on radia-
tive processes seemed to work very well, if one adopted the rule
that infinite radiative corrections should be regarded as small.
These successes persisted ~n a qualitative way with the intro-

duction of the Fermi theory of the weak interactions and the Yukawa
theory of mesons and nucleons. Calculated in Born approximation,
the Fermi interaction gave a semi-quantitative account of nuclear
S decay. (It was semi-quantitative rather than quantitative only
because the experiments were wrong; the experimental difficulties
with the S ray spectrometers were straightened out only in the
1940's.) The Yukawa theory gave a theoretical description of the
mesons which had been discovered and an attractive explanation of
the short range of nuclear forces, and nuclear S decay.
988 A. S. WIGHTMAN
Nevertheless t the occurrence of divergences in the theory was

taken as prima facie evidence of internal inconsistency, and the
typical response to this supposed inconsistency was to search for
possible radical physical alterations of the foundations. Here are
three illustrations of this attitude taken from the middle and late
thirties and early forties:
1) When the first indications of the anomalously great penetrating

power of cosmic radiation appeared, it was generally assumed that
quantum electrodynamics had broken down at high energies. Even
after the theory of electron-photon showers was developed and shown
to agree with experiment, and the penetrating component of cosmic
radiation shown to be mesons t scepticism about quantum field theory
persisted.
2) When Heisenberg pointed out the difference between theories

like quantum electrodynamics and the Yukawa theory of mesons in
which the coupling constant is dimensionless and theories like the
Fermi theory of S radioactivity in which the coupling constant has
the dimensions of a positive power of length t he coupled his general
discussion of the consequences of this distinction to a proposal
that a fundamental length be introduced into the theory of elementary
particles 5)
3) When Heisenberg launched his famous series of papers on the re-

lativistic S matrix, he viewed the formalism as one which would pro-
vide an alternative to Lagrangian field theory; an alternative
capable of incorporating such radical ideas as the occurrence of a
6)
fundamental length
DIVERGENCES AND PERTURBATlVE RENORMALIZATION THEORY
The key to understanding the then existing attitude of scepti-

cism towards quantum field theory is the divergences. Admittedly
the arguments which were used to show that divergences represent a
fundamental difficulty were usually rather sketchy. For example,
. . 5)
~n the well-known paper on the fundamental length ment~oned above
Heisenberg says: "If one applies the prescriptions of quantum the-
ory to a relativistically invariant wave theory in which interac-
tions of waves (i.e. non-linear terms in the wave equation) occur,
one obtains divergent results, as has been noticed many times.
What is involved is that relativistic invariance requires a 'theory
of local interaction' in which the interaction is restricted by the
requirement that the propagation velocity of a wave at a point is
determined by the amplitude of another wave at this point. Because
of the infinite number of degrees of freedom of the continuum, i.e.
because of the possibility of waves of arbitrarily small wave length,
however, the eigenvalues of a wave amplitude at a fixed point are
infinite. This contradiction -- which is rather similar to the con-
tradiction in the Rayleigh-Jeans law -- clearly does not really mean
that the relativistic wave theory or the quantum theory is false and
has to be improved but shows that when quantum theory and relativis-
tic wave theory are combined, a universal constant of the dimensions
of a length must be taken into account." Even if one regarded this
argument as unconvincing, it was impossible to ignore all those di-
vergent integrals.
Later on, the divergence difficulties were located more pre-

cisely. For example, at the end of the 1940's Van Hove showed 7 )
that if ~ is any vector in the domain of the free Hamiltonian and
~ and HIK are respectively the total Hamiltonian and the interac-
tion Hamiltonian of a cut-off theory with cut-off K, then
990 A. S. WIGHTMAN
This result does not exhaust the subject, but it shows how one can
give a precise meaning to the statement that the theory makes no
sense in the limit in which waves of arbitrarily high frequency are
allowed to interact.
Perturbative renormalization theory, which had its roots in

work of Dirac on charge renormalization in the 1930's and Kramers
and others on mass renormalization in the 1940's, was developed
systematically in the period 1945-52. It represented a decisive
step forward in the problem of divergences. It was shown that the
divergences occurring in the perturbation expansions of the Green's
functions of renormalizable and super-renormalizable theories can
all be expressed in terms of a finite number of basic divergent
quantities. Isolating these in the perturbation series and replac-
ing them by finite values is the process of renormalization. (For
our present purposes a theory may be called renormalizable if its
coupling constant is dimensionless, super-renormalizable if the
coupling constant has dimension a negative power of a length, non-
renormalizable if a positive power of a length.)
The first thing to be said about this procedure is that in

spin 1 quantum electrodynamics it has been a dazzling success. The
magnetic moments of electron and muon, the Lamb shift, the fine
structure of positronium, the hyperfine structure of hydrogen are
in agreement with theory to a very high order of accuracy. Although
there have been cloudy periods in which the agreement with experi-
ment was in doubt, these discrepancies have. up to the present day,
always disappeared when more precise experiments were completed or
errors in theoretical calculations were eliminated.
A natural question when one looks at some renormalized pertur-

bation series, say the one for the gyromagnetic ratio, g, of the
electron
(1)
"31-. 0 ...
is whether the series actually converges. The standard answer is

that although it has not been proved, it is likely that the series
is asymptotic. It is worth digressing to review the elementary
mathematical definitions and facts that lie behind this assertion
especially in view of recent attempts to "sum the series". The fol-
lowing discussion to some extent repeats that of B. Simon's Coral
Gables reviews), but the facts are probably worth repeating.
Given a function of f, defined on an open interval a < x < b

one says that a function fN defined on the interval is asymptotic
to f of order N at a if
o .
This general definition applies in particular to the expression

fN(x) = \N a (x - a)n. If one is given a sequence {ao,al,az t }
L.n=o n
and \N a (x - a)n is asymptotic to f of order N at a for all
l.n=o n
.
:t t h en one says that the ser~es
Loo a ( x - a )n _~s
.
n=o n
asymptotic to f at a and writes
f ( x) a.n. ()(-4)
V\.
.
992 A. S. WIGHTMAN
Now if lim [f(x) - fN(x)][x - a]-N = 0, it evidently follows that

x-+-a k
~~ [f(x) - fN(x)][x - aJr = 0 for k = 0,1, N - I, so when
, a (x - a)n is asymptotic to f of order N at a
Ln=o n
~ {C X) =: 40 ~ f (a,+ )
tl~et
t".wt.- [ f(x\-f(a.t)- f{a+)(x-~)] [ )(-a.] -2~a.2.= f t' /A+)

l""-
Jl.a... 2 !
Thus, when a function f is asymptotic of order N at a to an ex-

P ression Ln=o
,N a (x - a)n the function f necessarily has N deriva-
n
tives from the right at a and the coefficients a are uniquely
n
determined as Taylor coefficients
n.. = 0l.f I N (2)
Thus the assertion that the gyromagnetic anomaly g-2/2 has the asymp-
totic series
n
,00
a (r:J./'if)n at 0 means no more and no less than that
Ln=
1
g-2/2 is defined for r:J. in some interval 0 < r:J. < r:J.o and has deriva-
tives of all orders from the right at O. Then the a are uniquely
n
determined; (2) holds for all N with f = g-2/2 and a = O.
A second natural question is whether there are any restrictions

on the a which result from (2). An answer was given by Hadamard
n
in the nineteenth century. He showed that there are functions
differentiable of all orders on an open interval containing a (and

therefore having their Taylor series at a as asymptotic series)
with completely arbitrary values of a. There are no growth restric-
n
.
t~ons on f(n) ( a ) as a funct~on
. 0 f n even ~. f"~t ~s assume d that f ~s
.
differentiable of all orders in an open interval containing a. Of
course. this behaviour contrasts strongly with the behaviour of a
function analytic at a. By definition. f is analytic at a if it
is defined in an open interval containing a and there exists a
power series \00 a (x - a)n convergent in some open interval con-
Ln=o n
taining a. whose sum coincides with f in that neighbourhood. When
n
f is analytic at a. it obviously has \00 a (x - a) as an asymp-
Ln=o n
totic series at a with a given by (I), but there is also a growth
n
condition on the a
n
for some c. a bound which one can derive from Cauchy's integral for-
mula for the a
n
The divergence of the perturbation series for a quantity like

(2) in quantum electrodynamics is still not settled but in simpler
models analogous series are known to diverge. (The divergence of
the series should, of course. not be confused with the divergences
removed by renormalization theory.)
There is also an argument for the divergence of the series of

quantum electrodynamics found by Dyson early in the 1950's which de-
9)
serves some comment Dyson noted that if the series in a con-
verged in some neighbourhood of 0, then they would certainly converge
for negative a. He then interpreted the resulting Green's functions
with a negative as those of a theory in which like charges attract
and unlike repel. He argued that the vacuum of such a theory would
be unstable because states of zero charge lower than the vacuum would
994 A. S. WIGHTMAN
exist. To see this, note that a state with a large number of elec-
trons of like charge could be made for which the binding energy
would be arbitrarily large and negative. growing, in fact, like the
number of electrons squared. If this electron cloud were combined
with a like cloud of opposite sign of charge and far away, one would
have a state of zero charge lower than the vacuum. Dyson then ar-
gued that the occurrence of such a vacuum instability would not per-
mit sensible Green's functions to exist, and therefore the hypothe-
sis that there is analyticity in the perturbation series at 0 is in-
admissible. There are at least two objections which can be raised
against Dyson's argument. First, it is not obvious that the analy-
tic continuation of the Green's functions to negative a are the
Green's functions of a theory in which like particles attract. In
fact, the Hamiltonian of such a theory would be of the form Ho iH I ,
non-Hermitian. Secondly, just because the vacuum is unstable it is
not obvious that the Green's functions fail to exist.
The amazing thing about Dyson's argument is that it is likely

that both it and the second objection to it are right. We know this
for the analogous case of the anharmonic oscillator as a result of
the work of Bender and Wu and others. The situation there will be
discussed later. If one accepts the results, by analogy, for QED one
would have i) a cut on the negative real axis so a = 0 is not a
point of analyticity. ii) meaningful Green's functions for a =
= -Ial io, so the Green's functions for negative a do not blow up;
iii) a physical interpretation of the Green's functions for negative
a that does involve an unstable vacuum. If we accept this analogy
with the anharmonic oscillator, Dyson's argument seems to satisfy N.
Bohr's criterion for a really deep argument: it is both right and
wrong!
If a series is asymptotic, is it useful theoretically or experi-

mentally? If one knows nothing about the function to which the series
is asymptotic, the answer is no. On the other hand, the "practical"

attitude is this: if one knows that the a increase no worse than
n! then the nth term will be ~ 1 for n ~ 1/a, r:=o an an should be
better and better as a description of the function until one gets
to within an appreciable fraction of 137 th order. Then one should
stop; the remainder will be of the order of magnitude of the last
term kept. This may be right and it may be wrong; it remains to
be seen in such a series as (1). To be really sure one needs good
estimates of the remainder term in the Taylor formula with remain-
der
where
x
~
(>l-t:) Jt
On the theoretical side there are things that can be done if one
knows something about f. There are so-called summabi1ity methods.
These are operations on infinite series, divergent or convergent,
which yield convergent series or functions. Two elementary examples
are Cesaro summabi1ity and Abel summabi1ity. The first converts
i.e., it averages the partial sums. This operation yields ~ as the

sum of the series r:=o (_l)n. The second converts
996 A. S. WIGHTMAN
~ L"'" ttm, yo "'"

1"-').(- IVl=O
Applied to r:=O (_l)n it also yields %, because I:=o (_l)n rn = l/l+r.

A more powerful method of summability is that of Borel. Applied on a
formal power series
it yields
Here the sum under the integral sign is understood to be defined by

analytic continuation if necessary. I will discuss the application
of Borel summation to the perturbation series of quantum field theory
models later.
The three preceding summability methods are linear operations.

An important non-linear operation is that of Pade approximants. It
. 1 sum
rep 1aces the part1a IN a zn b y a rat10na
. 1 f unct10n
. p(z)
L M=
n=o n ,
= QL(z)/~(z), where QL and ~ are polynomials whose degrees, Land
M, respectively, add up to N. The coefficients of the polynomials
are adjusted so that the rational function has the same first N + 1
Taylor coefficients as the partial sum. This determines the polyno-
mials uniquely up to a common factor.
The existence of such summability methods suggests that one at-

tempt to obtain the non-perturbative solution of quantum field theo-
ries directly from their renormalized perturbation series by apply-
ing a summability method. Such a proposal raises the question of
uniqueness and the related question: if one were to define a pro-
posed solution of a Lagrangian quantum field theory as the result of
applying a summability method to its renormalized perturbation se-
ries, how could one be sure that the proposed solution actually
satisfies the equations of the theory? The straightforward answer

to the second question is that one would have to prove it directly.
On the other hand, if one had a priori knowledge that the summabi-
lity method would lead to a unique answer, one would have at least
a positive answer to the question: does the perturbation series de-
termine a non-perturbative solution?
If it were easy to sum the renormalized series explicitly us-

ing a summability method, it would, no doubt, be done just to see
what happens. However, it is very complicated in general, so one
is led to attempt to prove properties of the non-perturbative solu-
tions which will guarantee that a particular summability method will
necessarily yield a unique answer, the right one. A simple property
of this kind, whose significance has been emphasized by Simon is:
Definition
A formal power series ,00

L.n=o
a
n
zn is a strong asymptotic series
- -
for a function f(z) if and only if
i) f(z) is analytic in a region
for some and B > O.
ii) For some A and B and all N 0,1 and all z E R B

.
The first important fact about this notion is that a given formal
power series has at most one function to which it is strongly asymp-
totic. Second, Borel's summability method always works for strongly
asymptotic series. More explicitly:
998 A. S. WIGHTMAN
Theorem
Let F be analytic in a neighbourhood R B as defined above in

E, (
(i) and satisfy the estimate (ii), then f has derivatives f n)(O+),
of all orders n 0,1,2, from the right at the origin. Its for-
mal Taylor series has an associated series
0\\ ~(Y\.)
(D+) ~
[ 2
fl::O (~!)2.
which converges for Izl < B and defines an analytic function F which
may be continued analytically to the sector larg zl < E. f can be
recovered from F by the integral formula
A more special class of functions for which a somewhat more ex-

plicit recovery is possible is the following. Call a function
Stieltjes if it is analytic in the complex plane cut along the nega-
tive real axis -00 < X < 0, and can be represented
00
=
o
f .!..fi!!
t:+~
where dp is a positive measure satisfying

o
J tN\..df(tJ(Oo '\1. :: 0, f I 2.
Theorem
If f is Stieltjes, its diagonal Pade approximants PN N converge

in the cut plane to f uniformly on compact sets, provided lanl ~
n
~ A B (2n)! for some constants A and B.
The eigenvalues of the anharmonic oscillator have been shown

to be Stie1tjes as functions of the anharmonic coupling as will be
described later.
It is instructive to examine the example f(z) = exp (lIz a. ),

o< a. < 1 in the light of these theorems. It has the property
f(n)(O+) = 0, n = 0,1,2, It is left to the reader to show
that f violates the bound (ii) in the definition of strongly asymp-
totic and is not Stie1tjes. This kind of function is typical of
what has to be shown to be absent from the exact solution of a quan-
tum field theory, if the renorma1ized perturbation series is to de-
termine the solution.
All this information on summabi1ity methods was available in

the 1950's but apart from the special case of a quantized spinor
field in a constant external magnetic field where the explicit ex-
pression for the action was shown to be Borel summab1e, there ap-
pears to have been no attempt at that time to use summabi1ity
methods on the renorma1ized perturbation series of quantum field
theory. I will return to these methods later in connection with an
assessment of our current situation.
What was done in the 1950's was the development of non-pertur-

bative methods like dispersion theory and S matrix theory and their
application to nucleons and pions. The disentangling of the weak
interactions and the beginning of the analysis of symmetry breaking
also lent credibility to quantum field theory, but it was the de-
velopment of axiomatic field theory that contributed most to the
questions I am concerned with here.
From the point of view of the problem of divergences the most

important thing that the theory of perturbative renorma1ization did
was to make it plausible that counterterms cancelling the infinite
self-energies and charge renorma1izations had to appear in the
1000 A. S. WIGHTMAN
Hamiltonian in order to make it well defined. Looked at from a

conceptual point of view, the question of the existence of a non-
perturbative renormalization theory was reduced to a purely techni-
cal problem: can one introduce a cut-off dependence into the basic
quantities of the theory in such a way that the Green's functions
have a limit and define a field theory when the cut-offs are re-
moved? However, another step in the evolution of the subject was
necessary before this problem could be tackled.
THE IDEA OF A GENERAL THEORY OF QUANTIZED FIELDS
The motivation for the creation of an axiomatic field theory

was simple enough: it was frustration. Perturbative renormaliza-
tion theory is exceedingly complicated and was even more so in the
early 1950's. Everyone agreed that it should be generalized to be
non-perturbative. However, a simple-minded person asking: what ~s
the problem and what is to be regarded as an acceptable solution of

it? could get no answer from the available accounts of quantum field
theory. A general theory of quantized fields ought to contain de-
finitions of the notions of quantized field and quantum field theory
including a list of indispensable attributes. Such definitions would
at least permit one to define a notion of acceptable solution in
non-perturbative renormalization theory.
The list of axioms turned out to be agreeably short. but they

are so familiar that I will not write them down. Instead I will
comment on a few aspects of them that were regarded as somewhat no-
vel at the time. First, fields were defined as operator-valued dis-
tributions. That means that one gives meaning as an operator in the
Hilbert space of states to a quantity like
=
where f is some smooth test function although, in agreement with

what Heisenberg said in the quotation above, (x) itself is ill-
defined as an operator. Mathematically, this use of distributions
was an application of relatively recent ideas since L. Schwartz'
systematic book on the subject had just come out. In physics, the
use of such generalized functions in an unsystematic and heuristic
way had been standard for a long time. What was novel physically
was the use of a class of test functions to control the high-energy
behaviour of a theory. The assumption that appropriate matrix ele-
ments (f) are continuous linear functionals of f when f lies in
Schwartz' space of rapidly decreasing functions implies that the
vacuum expectation values,
tributions. With other "smoother" test functions worse high-energy
behaviour is allowed.
The second novelty about the theory was that it should exist
at all. Why should there be a general theory of quantized fields?
I think the answer ~s that the use of Green's functions on vacuum
expectation values to express the physical content of a theory (a
general idea for which J. Schwinger can claim as much credit as any-
body) is, on the one hand, very close to physical applications via
reduction formulas for S-matrix elements and, on the other hand, is
simply expressed in terms of the fields and physical vacuum. In
any case, it turned out that a rather impressive collection of gene-
ral results can be proved directly from the axioms without a detailed
use of the structure of equations of motion (PCT theorem, Spin-Sta-
tistics theorem, Haag-Ruelle scattering theory, some dispersion re-
lations, Goldstone theorem, etc.).
The successes of axiomatic field theory led naturally to the

question: is there a general structure theory of quantized fields?
Such a structure theory would be of the greatest importance both in
principle and practice because it would guarantee the general
1002 A. S. WIGHTMAN
requirements of Lorentz invariance, spectral condition, and locality

while displaying the arbitrariness to be attributed to detailed dy-
namics. Despite considerable effort, the search for such a struc-
ture theory was unsuccessful. It appeared to be premature; a de-
tailed study of examples would be necessary before there could be
much hope that the key concepts of the structure theory could be
located. This led to the program of constructive field theory as
will shortly be discussed.
A last comment on axiomatic field theory is in order. That

concerns its relation to quantum electrodynamics. In the three
standard books on axiomatic field theory, attention is restricted
to theories in which the Hilbert space metric is positive. This
restriction is natural on grounds of simplicity but it excludes QED
at least when QED is written in terms of a Dirac field and vector
potential in any covariant gauge. On the other hand, Gupta and
Bleuler showed how such covariant gauges can be written in terms of
a formalism in which the state space is enlarged to a triple of
nested spaces
with an indefinite scalar product <0,0>, positive on 1t' and vanish-

ing on 1-t". (1(.' and U" are supposed to be maximal wi th these pro-
perties.) The physical states are then described by vectors in
with - - - being the closure in the norm, 11 011 ~, of the quo-

tient space of +t, by iot". Physical observables leave~' and $ttl
invariant and so graduate to operators int pHyS ' Unobservable quan-
tities like the vector potential do not leave~' invariant and all
of 1t is necessary to give them a proper definition. This formalism
seems to have aroused nearly universal revulsion but it is really not

so bad once you get used to it. Anyone who can swallow a Faddeev-
Popov ghost should be able to swallow the indefinite metric formalism.
In fact, he has to swallow it because they are equivalent descrip-
tions.
CONSTRUCTIVE QUANTUM FIELD THEORY
The goal of constructive quantum field theory is simple I t is

to establish the existence of solutions to specific Lagrangian quantum
field theories. One starts with a cut-off version of the theory (for
which existence is easily established) and then studies the limit as
the cut-off is removed The strategy is to begin with the easiest mo-
dels and introduce the difficulties step by step. Typically, as time
goes on in addition to existence more refined properties are proved.
Of course, in keeping with the standards of mathematical physics, a
res,ult is regarded as a result in this comer of the forrest only if
it is stated and proved as a mathematical theorem. Somewhere along
the way, the axioms of axiomatic field theory are proved so the gene-
ral results of quantum field theory become available.
Here are some results of more than a decade of work by Glimm and
10)
Jaffe and their followers :
I Theory Axioms Symmetry

breaking
Existence of
single part
Borel surnmability
of
state Schwinger functions
A(4)4)2 ,I in one phase region For all sufficiently ,I ,I

large A
AY2 ,I for sufficiently small A Ys for all sufficient ly ,I ,I

large A
A(4)4)3 ,I for sufficiently small A With addition of 0(4)2)3 ,I ,I

o sufficiently negative
1004 A. S. WIGHTMAN
These models are all super-renormalizable but \Y2 and \(4)3

do have non-trivial divergences. The S matrices of all the models
are non-trivial. Thus, the question of the existence of non-trivial
relativistic theories in two and three dimensions is settled. These
examples make clear that if one wants to argue that divergences re-
sult in the inconsistency of relativistic quantum field theory, one
will have to find an argument which works in four dimensions but
fails in three and two.
It is interesting to see what can be said about analyticity in

the coupling constant and summability methods for the perturbation
series in these models. In fact, first let me discuss an even simp-
ler case, the anharmonic oscillator ground state energy Eo(a,S).
It is the lowest eigenvalue of the operator
defined in L (-00, 00; dx) for -00 < a < 00, 0 < S < 00. The expansion
in which we are interested is a power series in S
(~) ~Y\..
E"o (0<,0+)_
I\<\.!
There was surprisingly little known about Eo(a,S) and this expansion
until the late 1960's when the following facts were established,
Eo(a,S) satisfies the scaling relation
and since ax 2 is a regular perturbation of H(O,S), Eo(a,S) possesses

a convergent strong coupling expansion
- 2/'3 i\'\.
C"" (ol~ )
convergent for all sufficiently small laS

-2/3
I. Eo(a,S) has deriva-
tives of all orders in S from the right at O. Eo(a,S) is a Stieltjes
function and so the Pade approximauts of its perturbation series con-
verge to it. The nature of the singularity at S = 0 is indicated by
the diagram where the dashed lines indicate a family of branch cuts
on the second sheet with the origin as cluster point. So that i.s
what the essential singularity at zero couplings look like.
"".---- .... "

~ X' .... --- ... x
..
....
~
x--- ...x
Fig. 1 The essential singularity
X'----x at S = 0 for Eo(a,S)
x....... ___ ."",x x
~....
.......... _-_ .... ","
For A(~4)2 and AY2 the analogous questions have been studied
for the expansions in A of the Euclidean Green's functions. Notice
that these are the perturbation seri~es in the unrenormalized coupl-
11 12)
ing constant A. The authors' use a generalized version of the
above theorem on strongly asymptotic series which permits E =0 pro-
vided one has conditions on the boundary values of f(z) on the line
larg zl = rr/2.
Whether the Schwinger functions of these models are in fact,

analytic in the cut plane and have singularities analogous to those
occurring for the anharmonic oscillator is an open question.
I would like to make four further points about the results of

quantum field theory.
a) The founders of quantum field theory were right in not believ-

ing in quantum field theory (without renormalization!). In general,
1006 A. S. WIGHTMAN
these models have Schwinger functions which would blow up in the li-
mit in which ultra-violet cut-offs are removed if the renormalization
counterterms were not included in the Hamiltonian. Renormalization
is essential to make sense of these models.
b) The construction of the solution of A(~4)3 and Y2 was not only

a technical tour de force. It also removed a psychological barrier.
In a totally non-trivial theory, it was established that the counter-
terms really do cancel the infinities non-perturbatively.
c) The ingenuity and effort that went into the proofs of these re-
sults should not be underestimated. The proof of the existence of
the Euclidean Green's functions in A(~4)3 is the most difficult I
know of in mathematical physics. However, there ate also simple
general and enlightening ideas involved such as the method of cor-
relation inequalities.
d) Perturbation theory has shown itself very reliable as a guide

to the necessary renormalizations in all the models treated.
It is hard to resist saying more about constructive field theo-

ry, but I will do so and pass on to a subject which has given us
quite different insights into the foundations of quantum field theo-
ry.
LOCAL QUANTUM THEORY
In the 1960's there appeared a new approach to the foundations

of quantum field theory quite different from standard axiomatic
field theory. The basic objects of the new theory, which for brevi-
ty, I will call local quantum theory, are algebras of bounded opera-
tors (h(0) attached to regions 0 of space-time. The Hermitian op-
erators in ~(O) are supposed to be observables associated with the
region O. The G(O) are supposed to satisfy
a) 01C02implies <iI(Ol) ~~(02)
b) If the points of 01 are separated by space-like space-time

intervals from the points of O2 then 16;(01 ), ~(02)] O. The
algebra
Gt ::: U Gt (0) NO~~

o
where the bar signifies closure in operator norm. ~ is called the
quasi-local algebra and being a norm closed * algebra of operators,
. .
~t ~s a so-called C* algebra.
One of the important advantages of working with C* algebras is

that there is a nice connection between states on the algebra and
representations of the algebra. Here the word state is understood
as meaning a complex-valued linear functional on en satisfying
(positivity)
and
w (1) = 1. (normalization)
The representation associated with a state 11\ is obtained by the so-

called GNS construction. The result is that there exists a Hilbert
space,~ , a representation TI of (; by linear operators in~ , and
W W W
a distinguished vector ~ such that
w
The most important result of the local algebra formalism for

our present purposes is that in the hands of Haag, Kastler and
Doplicher, Haag,Roberts, it led to a theory of super-selection rules,
Let me sketch how this comes about using the electric charge super-
selection rule as an example. We know that in any theory that con-
tains a description of electric charge the Hilbert space is a direct
1008 A. S. WIGHTMAN
sum of subspaces of definite charge (sometimes referred to as super-

selection sectors or coherent subspaces)
it =
and observables leave each of the subspaces invariant. How are ob-
servations in one sector related to those in another? Local quan-
tum theory says that there is a quasi-local algebra of observables,
~, and that the representations of it in the various 4tne are unit-
ary inequivalent. Since the representations are also faithful (no
non-trivial observable is represented by 0) we can determine the
algebraic structure of ~ by looking at observations in anyone of
the sectors. What happens in the other sectors can then be con-
structed purely theoretically; one has just to construct the other
unitarily inequivalent representations of Gr. There is only one
proviso: not every representation is physically admissible. For ex-
ample, the representation may not lead to a unitary representation
of the Poincare group satisfying the spectral condition. A super-
selection sector is included in {t only if it is physically admis-
sible.
The importance of this general construction of super-selection

sectors has recently been underlined by discoveries in Lagrangian
field theory. There, after a vacuum sector had been constructed, it
was recognized that there might be topological quantum numbers, zero
in the vacuum sector, which would take values different from zero in
other sectors. Local quantum theory had already prepared a theore-
tical framework into which this process fits naturally.
Local quantum theory is important for another reason. Since it

works directly with observables, it makes possible a direct expres-
sion of the fundamental physical postulates of locality in space-
time, energy momentum spectrum condition, and Poincare invariance.
We will comment on this in more detail in connection with gauge

theories. On the other hand, Lagrangian field theory is much better
adapted to the construction of models and examples,
ASSESSMENT OF THE CURRENT SITUATION IN THE

MATHEMATICAL FOUNDATIONS OF QUANTUM FIELD THEORY
With all the above historical background in mind, let me turn

to the present and to the future. In accordance with the criterion
of belief in a theory stated above, I have first to assess the like-
lihood of internal consistency for quantum field theory in four-
dimensional space-time.
It is helpful to introduce here a diagram made popular in re-

cent years by work using the renormalization group.
d
dimension
of
space-time
9renorm
Fig. 2: The zeros of ~ as a function of the renormalized coupling

constant, g and space-time dimension d for A(~4)d
renorm
theory.
This diagram plots the zeros of the function S as function of re-

normalized coupling constant and the dimension of space-time. Of
course, we do not have space-times of fractional dimension available
but there are indications, E expansions around four dimensions, for
1010 A. S. WIGHTMAN
example, that interpolations between the integers exist. The hori-

zontal black lines at d = 1,2,3 indicate the range of coupling con-
stants for which constructive field theory has succeeded in construct-
ing solutions. The striking feature of the results is that as the
unrenormalized coupling constant varies 0 < A < 00 in (4)2 or (4)3,
the renormalized coupling constant remains bounded by what I have
denoted Wilson's zero. gw' If the principle of continuity makes any
sense it is the solutions for g > g that should be related to the
w
solutions of \(4)4' There are as yet no constructive results ~n
this region. All we have is the general guidance of the renormaliza-
tion group and the Glimm-Jaffe bound. The first says that the ultra-
violet behaviour of the theory should change as g crosses gw being
governed by the ultra-violet stable fixed point goo' which is the next
zero of B. The Glimm-Jaffe bound is a bound on g which can be inter~
preted as saying that the only way to get arbitrarily large g is to

have an unrenormalizable solution. It therefore makes plausible the
existence of g00 and g'.
00
(g00 determines the ultra-violet behaviour of
the renormalizable but not super-renormalizable solutions of the
(4)d model; g~ indicates the transition from renormalizable to non-
renormalizable solutions.)
Viewed from this perspective, the next great problem of construc-

tive field theory is to get past Wilson's zero. i.e., to construct re-
normalizable but not super-renormalizable solutions. The existence
of the renormalized perturbation series for (4)4 is good evidence
that such solutions should exist but a word of caution is in order.
If by poetic analogy we regard crossing Wilson's zero as a "phase
transition", in the (4)d model with d < 4, we may have to deal with
a phenomenon common in statistical mechanics. For the cut-off the-
ory, the basic quantities of the theory may be analytic in a neigh-
bourhood of g but as the cut-off is removed they may converge to
w
two quite different functions. (This analogy can be made quite pre-
cise if one compares the Euclidean expression for the vacuum energy
per unit volume in quantum field theory
Vi-'=> 60
with the grand canonical pressure 1n statistical mechanics.)
Do we have any evidence that this problem has no solution? Not

really. Of course, there is always Landau's ultimate argument. I
once expressed to Landau my lack of confidence in the arguments which
he and his co-workers had put forward for the inconsistency of field
theory. He then offered me the following: You agree that the es-
sential problem of quantum field theory is its high-energy behaviour?
Yes. You agree that up to now no-one has suggested a consistent high-
energy behaviour for quantum field theory? Yes. Then you have to be-
lieve in the inconsistency of quantum field theory, because physicists
are smart and if there was a consistent high-energy behaviour' they
would have found it!
I do not go along with Landau's argument. In fact, having con-

sulted the feeling in my bones, I am inclined to bet in favour of its
consistency. In fact, I will be more specific. In keeping with the
theoretical traditions of this School I would like to offer a theore-
tical prediction at the 5% confidence level: within five years, there
will be a rigorous construction of the solutions of A(~4)4 and of
spin ~ quantum electrodynamics in four-dimensional space-time. Of
course, there are important differences of technical detail between
theories such as the drastically different behaviour in the coupling
constant plane of asymptotically free theories as compared with A(~4)4
theory (see Fig. 3 and Ref. 13). As you have heard in 't Hooft's lec-
tures, because of the existence of these singularities, the perturba-
tion series of QeD cannot be Borel summable in the sense discussed
above. Nevertheless, it is of interest to study the Borel transform
of the perturbation series. As you have heard, one believes that the
1012 A. S. WIGHTMAN
------~~~------+_--~x
Fig. 3 The horn-shaped analyticity domain in the

complex coupling constant plane for asymp-
totically free theories. See 't Hooft's
lectures and Ref. 12.
resulting function has characteristic singularities at locations

fixed by the instantons and the required renormalizations. These
intriguing insights suggest the mathematical problem of finding
summability methods for functions analytic in a horn-shaped region
such as that of Fig. 3 and with appropriate growth restrictions on
deri vati ves .
It is worth emphasizing that the existence of such a horn for

the ground state energy, Eo(a,S), of the anharmonic oscillator when
a is negative is an immediate consequence of the scaling relation
dis cussed above
and the singularity structure for a > 0 shown in Fig. 1. The exist-
ence of these singularities shows that the usual Borel summability
methods do not work on the anharmonic oscillator when the potential
has a double minimum. On the other hand, for the A(~4)2 model in
the two-phase region Glimm, Jaffe, and Spencer have established the
analyticity of the Schwinger functions and vacuum energy per unit
volume in a sectorial neighbourhood {O < IAI < R. larg AI < E} 14)

Since this region has a non-zero opening angle. it is much larger
than the horn of Fig. 3. The contrast in this behaviour for the
anharmonic oscillator (space-time dimension 1) and A(4)2 model
(space-time dimension 2) is to be attributed to the vacuum tunnel-
ling which takes place for the former but not for the latter. Non-
Abelian gauge theories with asymptotic freedom are presumably ana-
logous to the anharmonic oscillator in this respect.
COMPARISON WITH EXPERIMENT; CONCLUSION
Having disposed of internal consistency. let me consider agree-

ment with experiment. For the quantum electrodynamics of muons and
electrons the agreement is spectacular. Also it is not a secret
that gauge theories of electromagnetic and weak interactions are
among the most exciting developments in particle theory of the last
few years. There are also the enticing prospects offered for strong
interaction theory by quantum chromodynamics. The detailed compari-
son of the predictions of these theories with experiment will obvi-
ously be a major industry for the next few years. I would like to
emphasize a different aspect which is also. broadly speakinga a com-
parison with experiment. Gauge theories provide a flexible language
for the description of experiment consistent with locality, the
spectrum, and relativistic invariance. The observables predicted by
these theories seem to obey the requirements of local quantum theory.
Putting aside the problems arising from black holes, we have no ex-
perimental evidence for the violation of the assumptions that go into
local quantum theory. This is a general experimental argument in
favour of relativistic quantum field theory.
I think that it is a reasonable bet that the second criterion

for be1ief is also satisfied and therefore we ought to believe in
relativistic quantum field theory.
1014 A. S. WIGHTMAN
REFERENCES
1) For a more detailed discussion see A.S. Wightman, Mathematical

Idealization and Theoretical Physics; The Role of Mathema-
tical Physics, Progress in Scientific Culture! (1976) 19.
2) M. Born, W. Heisenberg and P. Jordan, Zur Quantenmechanik II,

Z. fur Phys. ~ (1925) 606.
3) P.A.M. Dirac, The Quantum Theory of the Absorption and Emission

of Radiation, Proc. Royal Soc. London ~ (1927) 243.
4) W. Heisenberg, Uber die in der Theorie der Elementarteilchen

auftretende universelle Lange, Ann. der Physik (5) 32 (1938)
20.
5) W. Heisenberg, Z. fur Phys. 101 (1936) 533.
6) W. Heisenberg, Die "beobachtbaren Grossen" in der Theorie der

Elementarteilchen I, Z. fur Phys. 120 (1943) 513, II 120
(1943) 673.
7) L. Van Hove, Sur l'operateur Hamiltonian de deux champs quan-

tifies en interaction, Bulletin de l'Academie royale de
Belgique 1Z (1951) 1055.
8) B. Simon, Perturbation Theory and Coupling Constant Analyticity

in Two-Dimensional Field Theories in "Fundamental Interactions
in Physics and Astrophysics", p. 120, Eds. G. Iverson,
A. Perlmutter and S. Mintz. Plenum Press (1973).
9) F.J. Dyson, Divergence of Perturbation in Quantum Electrodyna-

mics, Phys. Rev. 85 (1952) 631.
10) For a general review, see the proceedings of the 1973 Erice
Mathematical Physics School, Constructive Quantum Field
Theory, Eds. G. Velo and A. Wightman, Lecture Notes in
Physics No. 25, Springer-Verlag, Berlin (1973).
11) J. Eckmann, J. Magnen and R. Seneor, Decay Properties and Borel

Summability for the Schwinger Functions in P(~)2 Theories,
Commun. Math. Phys. 39 (1975) 251.
12) J. Eckmann, H. Epstein and J. Frohlich, Asymptotic Perturbation

Expansion for the S-Matrix and the Definition of Time-Ordered
Functions in Relativistic Quantum Field Theory Models, Ann.
Inst. Henri Poincare XXV (1976) 1.
13) N.N. Khuri, Solutions of the Callan-Symanzik Equation in a

Complex Neighbourhood of Zero Coupling, Phys. Rev. D12
(1975) 2298.
14) J. Glimm, A. Jaffe and T. Spencer, A Convergent Expansion about

Mean Field Theory, I II Ann. of Physics 101 (1976) 610 and
631.
DIS C U S S ION
CHAIRMAN: A.S. Wightman

Scientific Secretaries: B. Jancewicz, F. Strocchi
DISCUSSION
DEO
You told us to believe in a physical theory if it is
internally consistent and has a fairly well defined area
of experiments for which it is accurate. I do not agree
that the second requirement is needed. Einstein once
said to the effect that the theory will predict what the
experiments will reveal. If theorists are guided solely
by experiments, we may have situations like the high
y-anomaly.
WIGHTMAN:
This is a comment rather than a question. I think
there are obvious reasons in favour of the second require-
ment as one wants to describe nature as it is.
DEO:
What is the predictive power of Axiomatic or
Constructive field theory?
WIGHTMAN:
There are general properties of QFT which have been
derived from first principles in the framework ofaxio-
matic field theory like the PCT theorem, connection be-
tween spin and statistics, dispersion relations, Goldstone
1017
1018 DISCUSSION
theorem etc. For constructive QFT, up to this point the

predictive power has been quite small. This is what al-
ways happens when people undertake to reexamine the theor~
from the foundations. If you look at the history of
modern Statistical Mechanics when a change took place at
the end of the fifties and people got more mathematical
about the foundations, the old timers were very unhappy
about the new developments. They said they were offens-
ively mathematical and that they were not solving any of
the problems of the subject. By the 1970's this was no
longer true and I think that an analogous process could
take place in QFT. For example it was not a priori
obvious that perturbation theory would be a good guide
to the non-perturbative solution even for the super-
renormalizable models. It turned out to be a good guide,
as proved in constructive field theory.
DEO:
Why historically were people only worried about the

ultra-violet problem and not about the infrared diver-
gences? What is the view point of axiomatic/constructive
field theorists?
WIGHTMAN:
People first chose the simplest models that they
could solve, and gradually complications were introduced.
Recently there has been a lot of work on the infrared
problem: first of all the work of Buchholz who showed
that there is a perfectly rigorous scattering theory for
massless particles. Also the work of Glimm and Jaffe on
the behaviour of QFT's at the critical point, where there
is no mass gap. There is a lot of work going on at the
moment about the infrared problem.
LIPKIN:
There is an interesting "interdisciplinary" corrunent
relating the lectures of Wightman, Gursey and Iachello
which also shows how the highly abstract work discussed
by Wightman has influenced bread and butter calculations
in particle physics. Both Arima and Iachello found SU(6)
syrrunetries in the real world of experimental data which
violated previous intuition. Iachello found a symmetry
in nuclear data suggesting a model of interacting bosons
DISCUSSION 1019
which obviously did not exist in the nucleus, but he was

daring enough to pursue this symmetry and leave the ex-
planation of it until later. Professor Gursey has mod-
estly skipped over the daring of proposing a hadron model
which violated .the well known relation between spin and
statistics and left its explanation until later. (Stud-
ents today do not appreciate how revolutionary this idea
was at the time, just as they do not realize that Gell-
Mann in his proposal of strangeness was also very revol-
utionary because bosons with half integral isospin were
believed at that time to be also in conflict with basic
principles.)
The SU(6) statistics paradox was resolved by
Greenberg's symmetric quark model based on parastatistics,
which we know today to be equivalent to the introduction
of three color degrees of freedom. But the algebra of
local observables was crucial in proving this equivalence.
Without this powerful abstract concept and proof, we would
have two competing quark models today, one with para-
statistics and one with colored quarks. They would look
formally very different, and calculations with both models
would be made in an attempt to find experimental tests
to choose between them.
Conventional attempts to prove the equivalence of
the two models would fail, because they are formally not
equivalent. The added ingredient that local observables
are relevant is essential. A simple example of this is
the statement that the observable properties of a hadron
on the earth should not be affected by the presence of a
hadron on the moon even though the statistics required
symmetrization or anti symmetrization of the two quark
wave functions.
WIGHTMAN:
I would like to add one comment about confinement.

It illustrates one of the remarkable features of current
gauge theories. To understand it, one has to contrast
the kind of behaviour that occurs there with the "con-
finement" occurring in solid state physics. In solid state
physics, there are substances for which there is approx-
imate confinement: insulators with very broad forbidden
bands. In these substances, to excite an electron up to
a free electron state requires a lot of energy. The low
lying excitations form a rich spectrum of bound states,
1020 DISCUSSION
excitons, etc. The breadth of the forbidden band and

hence the "degree of confinement" varies with the charge
on the nucleus. If there were an analogue of this kind
of confinement in particle physics, one would have the
following behaviour: for weak coupling the quarks would
be free, but as the coupling constant increased to some
critical value, the ratio of the quark to meson and bary-
on masses would approach infinity. For larger values
of the coupling there would be no quarks in the theory.
This behaviour should be contrasted to confinement in
gauge theories, where the quarks remain with finite mass,
and get pushed out into the indefinite metric space,
where they are not physically realizable states. The
local quantum theory of the observables of such a gauge
theory has to contain implicitly all this information!
PRESKILL:
I am curious about the curve you sketched which gave
a zero of the ~ -function as a function of the dimension
in the (~~)Q model. I wonder how explicitly this
function is known. Do you just have a bound to the left
of which the curve must lie, or do you have more precise
information?
WIGHTMAN:
The part of the curve which is best understood is

the part related to the ~ expansion which gives the
asymptotic form of the curve near four dimensions. Sev-
eral terms of the series are known. One also knows an
absolute bound (the Glimm-Jaffe bound) on the renormal-
ized coupling constant gren. I regard the G-J bound as
possibly being a bound on the zero which separates the
region of renormalizable theories from the region of non-
renormalizable ones. The G-J bound would be violated in
a non-renormalizable theory.
PRESKILL:
Do you have a heuristic understanding of the Glimm-
Jaffe bound?
DISCUSSION 1021
WIGHTMAN:
The occurrence of a bound is not new. Such bounds

were derived by Martin et ale under the assumption that
there are no bound states. No such assumption is requir-
ed for the G-J bound. To understand why there should be
such a bound consider the 2-dimensional case. There one
can plot 8ren vs the bare coupling constant. The maximum
is reached at the critical
point which separates the
one-phase region from the
two-Cor more)-phase region.
Here you don't get higher and
higher values of 8ren by
making the coupling A
stronger; instead you get a
phase transition. Since
?
c<'iticoIl poi,.,t
:cP"":= 4>4_ <~"&')4>2.
one has an effective potent-
ial with two deep wells. The two theories corresponding
to the two phases sit on the
bottom of each well and as
~ becomes large the two
wells become deeper and deeper
and further and further apart
so that they are better and
better approximated by harm-
onic oscillator potentials,
which means that the coupling
becomes effectively weak.
PRESKILL:
Since there is no known nontrivial fixed point in

the 4-dimensional models, it is conceivable that there
are no renormalizable theories between the superrenormal-
izable and nonrenormalizable sectors in your picture.
Do you have a good reason to believe that they are
there?
WIGHTMAN:
Apart from prejudice, I don't.

1022 DISCUSSION
KENWAY:
Many people draw graphs of the ~-function for all

values of the renormalized coupling constant. To what
extent is the ~ -function defined outside the range of
validity of the perturbation series?
WIGHTMAN:
In the context of ( 4>4)2 field theory the answer

is that the behaviour of the ~-function is known in the
range gren <
gw. Beyond the Wilson fixed point gw,
there is no construction of the theory. There could
well be a phase transition, as in statistical mechanics,
so that the system is described by completely different
functions for gren < gw and gren '> gw. In that case
we have no way of knowing the ~ -function for gren > gw.
In the case of the Sine-Gordon theory it is known that
the S-matrix is the same function for any gren and is
smooth at the fixed point. For (cp4)2 Glimm and Jaffe
have shown that an infinite charge renormalization is
necessary at the point g = gw, so that the situation
there is less clear.
KENWAY:
Could you comment on the relationship between the

existence of an ultra-violet fixed point and renormaliz-
ability of a field theory?
WIGHTMAN:
I can partly guess on the basis of Sine-Gordon field

theory where the interaction is of the form (c{/~) c.o~~c\:>.
There are three crucial ranges of the coupling constant
~. For 0 ~ ~~ ~ 41t the theory is "s uper" superrenorm-
alizable, in the sense that the perturbation series for
the Schwinger functions converges. In the range -4 1T ~
~ ~il ~ ~1t there are strong arguments that the theory re-
quires an extra renormalization. In the region ~2> ~1t
by using a variational procedure Coleman has shown that
the vacuum is unstable. Schroer and Truong regard that
computation unreliable because in this range you would
have nonrenormalizable solutions. This is an example of
how the renormalizability properties of a theory vary
with the coupling constant.
DISCUSSION 1023
The proper answer is given by the Glimm-Jaffe bound.

This tells you that if the theory has ~enormalizability
properties such that the integral f d P (a 2 )/a 2 for
the two point spectral function is convergent at 00,
then necessarily there will be a finite range for the
coupling constant.
DINE:
We have heard a good deal here about theoretical

work in Euclidean space, always justified by reference
to the Euclidicity postulate. In the models which have
been analysed so far by the constructive field theorists,
is anything known about the transition from Euclidean to
Minkowski space? Are there any interesting subtleties
or surprises? Is there any difference between massive
and massless theories?
WIGHTMAN:
From the practical point of view the most important

paper is that of Osterwalder and Schrader (there was
also an older work by Nelson which was very important
historically). The theorem is quite strong: if the
Euclidean n-point functions satisfy certain n-dependent
bounds the process of analytic continuation can be
carried out to get a Minkowski theory satisfying the
axioms of QFT. This holds in any number of dimensions.
CELMASTER:
Does the Osterwalder-Schrader theory apply to

fermions?
WIGHTMAN:
Yes. However the analogs of the techniques of LP

spaces for establishing the bounds on the functional
integrals are poorly developed when one is dealing with
Euclidean Fermi fields.
1024 DISCUSSION
TAO:
't Hooft derivation of bubble-like analytic

structure in the g2 plane was based on two assumptions:
a) Asymptotic freedom;
b) There is only one zero of the ~-function.
Did you say that there are exactly soluble models which
satisfy the two above assumptions and which are Borel
surnrnable? Does it mean that Gerard is wrong?
WIGHTMAN:
I have the impression that two facts could play a

role: the theory 't Hooft is discussing has no mass
scale built into it, moreover the diagrams I draw for
2-dimensional models refer to the unrenormalized coupling
constant. Certainly, analy-
tici ty in A has been establ-
ished for the region in the
figure for (c:PIf )2, and l'( ) 2.
theories, and on the face of
it the region looks differ-
ent from that obtained by
't Hooft. On the other hand
for the (~~)1 theory (the
anharmonic oscillator), the
work of Barry Simon shows the
existence of a horn of singularities like that discussed
by 't Hooft, but only if the
quadratic term in the Hamil-
tonian has a negative co-
efficient. (This was noticed
by A. Sokol.) The physical
......
.. "
1(
reason is simple. The

theory with a double minimum
potential predicts instantons,
but with a single minimum it
does not. The (c}+) 2. and
(~4)3 theories do not
predict instantons. As 't Hooft has explained, Borel
surrunability is not expected when instantons are present.
BACE:
You stated that QED might eventually be under contro]

If the electromagnetic a were about 1/3, 't Hooft's
DISCUSSION 1025
program would be very relevant for computations in QED.

Does "under control" mean that 't Hooft's program might
soon become superfluous for QED, i.e. that the rigorous
methods would allow computation?
WIGHTMAN:
One would hope so. Let me give a~ example. In sett-

ling the question of existence of ($)2. theory, Glimm,
Jaffe and Spencer gave an asymptotic expansion in 1/A
valid in the region of large ~ Every term of it can
be calculated with Feynman diagrams and it becomes better
and better the stronger the coupling. It also has an
error estimate. This kind of thing is eminently practical
if anybody cares about the strong coupling limit of (~~)
z.
MIETTINEN:
QED is one of the most beautiful and accurate phy-

sical theories known to man. At the same time, however,
one knows that the photon has a hadronic component with
a probability N ClC.= 1/137 N 1%. This means that QED is
not a closed and isolated field of physics but is intim-
ately linked to strong interactions. Now axiomatic field
theorists have worked hard to clarify the theoretical
foundations of QFT and to formulate a sound and self-
consistent theory for electromagnetic interactions. Do
you think that QED can be made self-consistent without
considering other than electromagnetic interactions?
WIGHTMAN:
My guess would be that QED will be an idealized

model, just as potential scattering. Its internal
consistency will be determined by high energy behaviour
in the region where if you really want to make a full
physical theory you have to take into account the whole
elementary particle spectrum. So the internal consist-
ency of QED is not a physical question. However, for
the clarification of ideas, understanding a limit of
idealized models is absolutely vital.
AN EXACT RELATIVISTIC S-MATRIX IN 1 + 1 DIMENSIONS:
THE ON-SHELL SOLUTION OF THE MASSIVE THIRRING MODEL
AND THE QUANTUM SINE-GORDON EQUATION
M. Karowski
Institut fur Theoretische Physik
Freie Universitat Berlin, Germany
1. INTRODUCTION
1.1 The results 1,2)
We shall calculate an S-matrix element like
<t (P') f (,.) .... L~

DU.t -
-l .... 1 (~ ).n. -j ('I) ... bj
(')
5 Il}
('ft. " . >
U\..
for the scattering of arbitrary numbers of particles (fermions) f,

antiparticles (antifermions) f, and different kinds of bound states
bl , b ,where b. can either be considered as abound state of (ff)
,
n 1.
or (b., b. .) or (b., bk , b. . k)' etc.
) 1.-J J 1.-J-
We shall see that the S-matrix in 1 + 1 dimensions is uniquely

determined by some general properties which are valid, e.g., in the
case of the massive Thirring model and the Sine-Gordon equation.
1.2 The Sine-Gordon equation
Let us consider the relativistic wave equation in 1 + 1 dimen-

sions:
1027
1028 M. KAROWSKI
o 4> +
where ~(x,t) is a classical field, va a mass parameter, and Sa

coupling constant. This equation is completely soluble by means of
. 3)
the inverse scatter~ng method There is a static solution called
"soliton"
<lis (X) =4/~ tg- 1 e'rax
-------------------------------
~===========---+_--~======~--~x
with energy or mass m = SIa/S 2 and the antisoliton A(x) = S(-x).

Another set of solutions, oscillating in time, are the "breathers"
which may be interpreted as soliton-antisoliton bound states. They
have a continuous (centre-of-mass) energy spectrum with 0 < E < 2m.
Now we come to the most important property of the Sine-Gordon

equation: there exists an infinite sequence of local conservation
laws 4)
I'V\ == -f1~.S,,,,
in
The corresponding charges of asymptotic states !Pl, . , PN>out are
AN EXACT RELATIVISTIC S-MATRIX IN 1+1 DIMENSIONS 1029
where P = Po Pl' (Similar currents exist for p_,) The conserva-

tion of these charges implies for scattering
i) no particle production;
ii) only momentum exchange, i.e., the sets of incoming and outgoing
momenta are equal;
iii) factorization of an N-particle S-matrix element into two-
particle ones:
(tJ)
S ( ?-1 I ~~
I )
(Since the factors on the right-hand side do not commute in general

the ordering has to be specified, see below.)
The usual quantization of a field theory - by defining Green's

functions in renormalized perturbation theory by means of the
Gell-Mann-Low expansion - is not very useful for the Sine-Gordon
equation if one wants to describe solitons. Since ~S is very far
away from the vacuum ~ = 0, which would be the starting point for
ordinary perturbation theory. The quantization of classical solu-
tions can be attacked by semiclassical methods S ). Dashen.
Hasslacher, and Neveu obtained in WKB approximation the bound state
spectrum which is now discrete, of course,
1t=-t,2., .... <,A.
where
1030 M. KAROWSKI
and
~l.
( -1 - 8'f'C )
is the soliton mass. This spectrum was claimed to be exact.

6)
Coleman gave arguments based on perturbation theory that
the quantized Sine-Gordon theory must be equivalent to another field
theory in 1 + 1 dimensions. the massive Thirring model. Other ap-
7)
proaches to this problem were presented
1.3 The Thirring model
The (massless) Thirring mode1 8 ) describes a self-interacting

Dirac field in 1 + 1 dimensions:
h
were J~ = ;;,).l ,I,
0/ yo/an
dg '~s a coup l'~ng constant. The operator solu-
tion of this model is known since a long while and the Wightman
functions
can be calculated 8 ). This field theory is. however, rather simple.

Because of infra-red effects there are no one-particle poles in mo-
mentum space and there is no scattering in this model. The massive
Thirring model defined by the Lagrangian
is much more complicated. Coleman's equivalence now says: identify

the two fundamental Thirring particles f, f with the soliton.
soliton, antisoliton of the Sine-Gordon equation and relate the

coupling constants of both models by
Our approach is, in some sense, an alternative proof of this equi-

valence and also a proof of the exactness of the WKB Sine-Gordon
spectrum.
2. THE MASSIVE THIRRING MODEL
Our treatment of the massive Thirring model starts with the

observation that for the classical case there exists again an in-
. . . 9)
f~n~te sequence of local conservat~on laws :
"" = .(,'3,S, ...
The fields ~, are anticommuting objects in the classical model;

we can consider it as the tree approximation of the corresponding
quantized model. The consequences of these conservation laws are
the same as for the Sine-Gordon equation mentioned above. This can
be seen as follows: a~
j~ = 0 holds true for solutions of the field
n
equation
In general, we have
= tl,.. F + iJ.\.c::..
1032 M. KAROWSKI
where ~ is a local operator which may be expressed by powers of

n
derivatives of the fields. The integrated Ward identity with X =
= ~(Xl) ~(~) in tree approximation reads
J <.,- ~r J: (){) x' >tr

o = J 2.)(
= JJ 2,., (.,- ( A Il F +- k. ~. J .x >tr

:;- (d 2){'
J
< T ( A tl (x ) ~ -4-
; 'rCx)
+ ~. c:: .) >( >
tr
=i (-rAn (X-1) ... 9(.~~,tt" + ... + l (T't'(x,,) ... At'. (x"')~r
In momentum space on the mass shell after amputation only the parts
in ~ contribute which are linear in the fields. These vanish for
n
even n and are given for odd n by
+ _.
where 13+ do + 13 1 Applying the LSZ reduction technique we ob-

tain
This means an S-matrix element vanishes unless l~ (P + i)n t n =

= 1,3,5, . is conserved. It follows that the sets of incoming
and outgoing momenta are equal
Hence we have the properties:
i) no particle production;
ii) only momentum exchange; and
iii) factorization of S.
It is believed 10 ) that (i) and (ii) imply (iii).
The question now is whether the conservation laws survive quan-

tization. The apparent occurrence of anomalies may be seen as fol-
lows. In the BPHZ scheme the integrated Ward identity reads 11 )
0:- Jd 2 x <T~t1[~~Jtlc)() X)
:: Jd x <T (~~ [At-. F] ()t. ) + h. c) X >

1
::Jce-x <T (tJJ - 2. [Alt] (x) ~ b 'f~

2. (x)
+ h.(:.) X>
_ Jd1x<r(N,,( [(~-[AK~)~otf](>') -4- k-C.) X)
The dimens ion d =n + 2 of the operator d

~
j~ determines the sub-
n
traction degree of the normal product Nd [d ~
j~J (x).
n
The unrenor-
mali zed mass mo differs from m by a counter term which is finite in
the BPHZ scheme and is determined by normalization conditions. The
reason for the extra term on the right-hand side is that moW in F
has lower dimension. Hence. in the quantum equation of motion in-
side of a normal product. Nd [Zn mow] has to be replaced by the ani-
sotropic normal product Nd [{En} mow]. This means that subgraphs
which contain the line ware subtracted according to d and those
which do not contain the line ware subtracted minimally (d - 1).
Therefore. the extra term will, in general. be different from zero
if 6 is not linear in the fields which is true for n ~ 3. One
n
1034 M. KAROWSKI
12) ]..I
expects that the currents j can be redefined such that these
n
anomalies cancel. This was explicitly shown for n =3 13). We be-
lieve that the conservation laws and the properties (i), (ii), and
(iii) hold true in the quantized massive Thirring model.
3. THE S-MATRIX
Our main result can be expressed as follows:
Theorem
If a relativistic model in 1 + 1 dimensions with a partic1e-

antiparticle pair f, f fulfills the assumptions:
1) i) no particle production
ii) only momentum exchange
iii) factorization of the S-matrix;
2) unitarity, crossing symmetry and T invariance.
3) non-vanishing backward particle-antiparticle scattering (and
some more technical properties concerning analytic and asympto-
tic behaviour which will show up in the proof),
then the S-matrix is uniquely determined. It depends on two para

meters: the mass m and a "coupling" constant A.
Note that assumption (1) is a consequence of the conservation

laws considered in the previous section.
Proof
An S-matrix obeying (1) can be formulated as fo110ws 2 ):

where a. denote the different kinds of particles f, f, b 1 , b 2 ,

1
The N-particle S-matrix is a product of all two-particle ones in a
special order which, for example, for PI > > PN is given by:
"'-1
~ (.~
~=i '~l ~W
where the two-particle S-operator is defined by
The transmission and the reflection amplitudes are t and r, respec-

tively. Because of (ii), reflection only appears for different
particles with equal mass, as f and f. Let us consider first "re-
pulsive" coupling (g < 0, S2 > 41T) such that there are no bound
states. The general case will be obtained by analytic continuation
with respect to the coupling. Then S(2) contains four functions:
t ff , t ff , tft: =t and r ff : = r.
Let us introduce the rapidity difference 6 12 by
The physical plane is mapped onto the strip 0 < 1m 6 < 1T and the
physical region for ff scattering (P1 + P2)2 - i > 4m2 corresponds
to 6 12 > O.
1036 M. KAROWSKI
itt
,---- - ---- - --- ------- - ---1
,
I I
I
T
. ...-.. . .
I
I C+I
=~---E===
o 4 m2 ~ L
I ___________ _ - - - - - - - ______1
o
Unitarity s+s = s(-e) see) =1 and crossing symmetry
r(9):. r(i.'ft-e)
yield
Hence, there remains only one unknown function tee) which we shall
now determine 1 ). If tn tee) is analytic in the physical strip and
does not grow terribly for IRe el + e, Cauchy's formula yields
-r
~ l(e) = -
1
--
C being the contour enclosing the physical strip 0 < 1m 8 < TI. The
unitarity and crossing relations imply
where h(z) = t(z)/r(z) which is well defined if r + O. We shall

now prove that h(z) is determined by the factorization condition of
the three-particle S-matrix and T-invariance: S(3) has to be sym-
metric: S12S13S23 = S23S13S12. This means, for example, that the
amplitude for the transition f(P1)f(P2)f(P3) ~ f(Pl)f(P2)f(P3) has
to be equal to that one of the reversed process
Using t(8) = r(8)h(8) = t ff (8)h(8)/h(iTI - 8), we obtain the func-

tional equation for h(8)
The solutions of this equation are
where A is a free parameter to be interpreted later. After some

calculations we obtain:
1038 M. KAROWSKI
'"
qe) = It)(P ~ J~)C
o
or
l(&) =- F(eljF (l1l'i-e)
with
This amplitude was first proposed by Zamolodchikovl~) who used the

extra assumptions (a) exactness of the WKB spectrum; (b) absence
of resonances. and (c) for integer values of A: absence of reflec-
tion and validity of a formula for t(S) due to Korepin and
Faddeev 1S ). In our treatment. the properties (a)-(c) follow from
the assumptions of the theorem.
The poles of the transmission amplitude t(S) in the physical

strip
k-=." .... ~>-.
correspond to (ff) bound states with masses
This spectrum coincides with the one calculated in WKB approxima-

tion for the Sine-Gordon equationS) if we relate the free parameter
A to the coupling constants Sand g by:
- ;( = 1 +
2.~
For A < 1, where there are no bound states, the S-matrix is com-
pletely determined. For A > 1 we have to calculate the scattering
of bound states, i.e., the amplitudes tbkf and tbkbi 2,16). We con-
sider the residue of S(3)(Pl,P2,P3) at (PI + P2)2 = ~ and obtain
after some calculations in agreement with previous results in the
' 1 aSS1ca
sem1C ' 1 I'1m1t
, 15) :
where
correspondingly for k ~ i:
where
So, finally, we can calculate the general S-matrix element written

down in the Introduction.
1040 M. KAROWSKI
4. OFF-SHELL QUANTITIES
17)
Using Watson's theorem , one can calculate the soliton form
IS)
factor
where
. :2.
4lM
>,.x..
--
n.
1:7
211:"""
o
In principle, matrix elements like
can be calculated by means of a generalization of Watson's theorem

and methods similar to those used for the derivation of the S-matrix.
This program, however, turns out to be rather complicated. The de-
termination of Wightman functions like
~s still an open problem.
The latter must have the same small distance behaviour as

those for the massless Thirring models) since the Callan-Symanzik
19)
function S in the massive Thirring model vanishes identically
ACKNOWLEDGEMENTS
This talk is based on collaborations with members of the In-

stitut fur Theoretische Physik, Freie Universitat Berlin: B. Berg,
S. Meyer, B. Schroer, R. Seiler, H.J. Thun, T.T. Truong and
P.H. Weisz.
REFERENCES
1) M. Karowski, H.J. Thun, T.T. Truong and P. Weisz, Phys. Letters

67B (1977) 321.
2) M. Karowski and H.J. Thun, Complete S-matrix of the massive

Thirring model, FU Berlin preprint 77/6, to be published.
3) M.S. Ablowitz, O.J. Kaup, A.C. Newell and N. Segur, Phys. Rev.
Letters 31 (1973) 125.
L.A. Takhtadzhyan and L.D. Faddeev, Theor.Math.Phys. ~ (1975)
1046.
For a review, see: A.C. Scott, F.Y.F. Chu and D.W. McLaughlin,
Proc. IEEE ~ (1973) 1443.
4) M.D. Kruskal and D. Wiley, American Math. Soc. Summer Seminar

on non-linear wave motion, ed. A.C. Newell, Potsdam, N.Y.
July (1972).
H. Steudel, Annalen der Physik 32 (1975) 205; cf. also:
R. Flume, Phys. Letters 62B (1976) 93 and corrigendum to be
published.
5) V.E. Korepin, L.D. Faddeev, Theor. Math. Phys. 25 (1975) 1039.

F.F. Daschen, B. Hasslacher and A. Neveu, Phys.~ev. Q!Q (1974)
4114, 4130, 4138; Qll (1975) 3424.
6) S. Coleman, Phys. Rev. !! (1975) 2088.
7) S. Mandelstam, Phys. Rev. Dll (1975) 3027.

R. Seiler and D. Uhlenbrock, Les Methodes Mathematiques de la
Theorie Quantique des Champes, No. 248 p. 363. Ann. Phys.
105 (1977) 81.
B. Schroer and T.T. Truong, Phys. Rev. D15 (1977) 1684.
8) W. Thirring, Ann. Phys. 3 (1958) 91.

V. Glaser, Nuovo Cimento-9 (1958) 990.
K. Johnson, Nuovo Cimento-20 (1961) 773.
B. Klaiber, Lectures in theoretical physics (Gordon and
Breach, New York, 1968) p. 141.
9) B. Berg, M. Karowski and H.J. Thun, Phys. Letters 62B (1976)

187; Phys. Letters 64B (1976) 286.
B. Yoon, Phys. Rev. D13~976) 3440.
R. Flume, D.K. Mitter-;nd N. Papanicolaou, Phys. Letters 64B
(1976) 289.
P.P. Kulish and E.R. Nissimov, Pisma v. JETP 24 (1976) 247.
1042 M.KAROWSKI
10) P.P. Kulish, Theor. Math. Phys. 26 (1976) 132.

R. Flume, V. Glaser and D. Iago1nitzer, unpublished.
11) W. Zimmermann, Ann. of Phys. 77 (1973) 536.

M. Gomes and J.R. Lowenstein,-Phys. Rev. D7 (1973) 550.
12) P.P. Kulish and E.R. Nissimov, Pisma v. JETP 24 (1976) 247.
13) The first proof for this conservation law was done in loop
approximation:
B. Berg, M. Karowski and R.J. Thun, Phys. Letters 62B (1976)
63.
For a general treatment, see: R. Flume and S. Meyer, Nuovo
Cimento Letters 18 (1977) 236.
B. Berg, M. Karowskr-and R.J. Thun, Nuovo Cimento 38A (1977)
11.
14) A.B. Zamo1odchikov, Moscow preprint ITEP-148 (1976). This

amplitude is in agreement with results of perturbative
calculations in the Thirring model coupling constant g and
in the semiclassical limit S + O.
P.R. Weisz, Nuclear Phys. B122 (1977) 1.
R. Jackiw and G. Woo, Phys:-R;v. D12 (1975) 1643.
15) V.E. Korepin and L.D. Faddeev, Theor. Math. Phys. 25 (1975)
1039.
J.L. Gervais and A. Jevicki, Nuclear Phys. B11D (1975) 113.
M.T. Jaekel, Nuclear Phys. Bl18 (1977) 508.
16) A.B. Zamo1odchikov, Moscow preprint lTEP-12 (1977).
17) K.M. Watson, Phys. Rev. 95 (1954) 228.
18) P.R. Weisz, Phys. Letters 67B (1977) 179.
19) M. Gomes and J.R. Lowenstein, Nuclear Phys. B45 (1972) 252.
DYNAMICAL S~lliTRIES IN NUCLEAR PHYSICS
F. Iachello
Kernfysisch Versneller Instituut
University of Groningen, The Netherlands
1. INTRODUCTION
The purpose of my talk is to report on some developments which

have occurred in the last two years in the field of nuclear physics
and to point out some similarities and differences between these
developments and those which have occurred in the field of elemen-
tary particle physics.
As is well known, one of the most important tools in cata-

loguing, classifying, and describing the properties of quantum
systems, is group theory. This tool has been used with some suc-
cess in elementary particle physics, as well as in atomic and mole-
cular physics. Applications to nuclear physics have been quite
scarce and limited to a small number of nuclei. At the very be-
ginning of nuclear physics, great importance was attached to
Wigner supermultiplet theoryl). Wigner observed that the spin
and isospin variables of the nucleons span a four-dimensional space
and proposed a description of nuclear levels in terms of the group
SU(4). This theory was abandoned after the discovery that nuclear
forces are not invariant under SU(4), because they contain strong
+ +
Heisenberg Tl T2 forces. Another theory which makes use of
1043
1044 F.IACHELLQ
group theory is that of Elliott 2 ). Elliott observed that if nu-

cleons move in a harmonic oscillator potential, then the single
particle levels are Is; lp; 2s, ld; 2p, If; .. , where the levels
between successive semicolons are degenerate in energy, and there-
fore, strongly mixed by residual interactions. For example, the
(2s) and (ld) orbitals are expected to mix strongly in nuclei hav-
ing 16 < A < 40, etc. Elliott also noted that the oscillator
Hamiltonian Ho = r2 + b 4 p2 is invariant, not only with respect
to rotations, but also with respect to the more general group
SU(3). He thus proposed a description of nuclear spectra in terms
of this group. Although Elliott's model has played, and still
plays, a very important role In nuclear physics, its applicability
is limited to the lower part of the s-d shell and it includes only
a few nuclei. This is because the spin-orbit interaction ~ i
removes the degeneracy of the single particle oscillator levels,
which is a necessary condition for having Elliott SU(3) symmetry.
For example, the ldo/2 and ld 12 levels are separated by ~ 3 MeV.
This problem becomes more and more acute as the nucleus becomes
heavier and heavier, and it makes Elliott SU(3) impractical for
A > 40.
As a consequence, it has usually been believed that methods

of group theory are not useful in nuclear physics, due to the com-
plexity of the many-body system and to the complicated form of the
nucleon-nucleon interaction.
2. INTERACTING BOSONS
This situation has been changing in the last few years. The
idea is that one is actually not interested in all the levels of
the many-body system but only in a small fraction of them, those
having some specific structure. For instance, the nucleus 156 Gd
may have 10 9 levels within the first 6 MeV of excitation, but we

are only interested in, say, 100 of them, the collective, discrete
states. Although the full many-body system may not have any
DYNAMICAL SYMMETRIES IN NUCLEAR PHYSICS 1045
symmetry at all, the subspace B of, for example, the collective

levels, may have some symmetries and these may be very useful in
classifying, cataloguing, and describing these states.
Thus the line of approach is that of replacing the complex

many-body system by a simpler system and then studying the sym-
metry properties of this system. This method has been successfully
applied to the study of the collective states in heavy even-even
nuclei 3 ,4). In these nuclei, a treatment in terms of the nucleon
(fermion) degrees of freedom is practically unmanageable, even
with modern day computers. However, it is well known that the ef-
fective nucleon-nucleon interaction between identical nucleons in
the same shell is strongly attractive in the state with total an-
gular momentum L o. That is, the two nucleons have a strong
tendency to form a state identical to a Cooper pair in the electron
gas. In addition, it is also well known that the effective inter-
action between identical nucleons in the same shell is also attrac-
tive in the state with total angular momentum L = 2, while being
zero or slightly repulsive in the other states. Suppose now that
we treat even-even nuclei not in terms of all the fermion degrees
of freedom, but only of L = 0 and L = 2 coupled pairs and that we
replace these pairs by bosons. In this way we have reduced the
complex many-body system to a system of interacting bosons, which
will be called s(L = 0) and deL = 2), respectively, as shown in
Fig. 1. To this simplified system, group theoretical methods can
now be easily applied, since one immediately notes that the five
components ~ = 2, +1, 0, -1, -2 of the L = 2 state, the d-boson,
and the single component of the L = 0 state, the s-boson, span a
s~x dimensional space, yielding SU(6), the special unitary group
in six dimensions, as the group structure of the problem. The
Hamiltonian H which includes one boson term and boson-boson in-
teractions
(2.1)
1046 F.IACHELLO
x )( X d L=2
x x x x x s L=O
Fig. I The configuration sSd 3 in the SU(6) boson model.

Es and Ed denote the single particle energies of
the L = 0 and L = 2 bosons, respectively.
can then be written in terms o~ the 36 generators g.1 of the grouD-

U(6). These r;enerators can easily be written dO\-ffi by introducing
boson creation (dt,st) and annihilation operators (d,s). They are
(2.2)
\vhere the parentheses denote angular momentum couplinr,s.
The single boson terms in the Hamiltonian are the scalars in

(2.2) while the two-boson interactions are the scalar products of
operators in (2.2) all multiplied by appropriate coefficients~
3. DYNAMICAL SYMMETRIES
The statement that SU(6) is the gorup structure of this prob-

lem is important in many respects. First of all, if this statement
is correct, the SU(6) space must be the representation space for all
wave functions of collective character and thus one should be able
to describe within it the large variety of observed spectra. These

spect~a are usually discussed in terms of the geometrical model of
Bohr and Mottelson 5 ) , wherein one imagines that the nucleus behaves
like a liquid drop with radius R = Ro (1 + I fl a
fl
Y2 (~)).
fl
In this
model there are two limiting situations which can be discussed
easily, that corresponding to vibrations of the drop around a sphe-
rical shape, and that corresponding to rigid rotations of the drop
around a deformed shape. Are these analogous situations in the
SU(6) model and can one establish a correspondence between the SU(6)
model and the geometrical description? The answer is yes. Not only
are there limiting situations in the SU(6) model analogous to those
of the classical geometrical model, but those limiting situations
corresponding to dynamical symmetries of the Hamiltonian H. Thus
group theoretical methods can be used in classifying the states,
in determining selection rules governing electromagnetic transi-
tions between states, and in providing mass formulas, etc.; in
other words in providing a complete and simple solution to the pro-
blem. In this respect, the geometrical and group theoretical de-
scriptions are complementary in the sense that the geometrical ana-
logue gives physical significance to the classification scheme,
which otherwise would merely be a mathematical device, while the
group theoretical description gives an elegant and concise descrip-
tion of the geometrical problem.
There are three dynamical symmetries which have been uncovered

by studying the group structure of the interacting boson model.
They can be labelled by the subgroup G of SU(6) which characterises
them and/or by the geometrical limit which corresponds to them:
i) G - SU(5), the anharmonic vibrator;

ii) G - SU(3) , the axial rotor;
iii) G - 0(6), the triaxial rotor with y = 30 0
The concept of a dynamical symmetry was introduced by Gell-Hann in

the context of particle physics. In group theoretical language it
1048 F.IACHELLO
means that the different representations of the group G are split

but not admixed by the Hamiltonian H. As I mentioned above, the
existence of these symmetries provides simple and complete solu-
tions to the problem. For instance, formulas can be written for
the energy levels of the system. I quote here, without derivation,
'+ )
the formula corresponding to the three cases g~ven above .
i) G == SU(S)
. ([ N] Y) d V 11 A L M) = EoJ -t 0( i nd (1'\ a- 1) +
(3.1)
f(nd-v)(nd+v+3)+~ [L(L+1)-6I\d]
ii) G - SU(3)
E([N] O'.r) I<LI1) = (3.2)
= (~ K-K')L(L+1)- k. [A2.+ tt 2+ Atl +3 (A-Itt)]
iii) G _ 0(6)
(3.3)
The symbols in parentheses following E on the left-hand side are the

quantum numbers needed to specify uniquely the states and the vari-
ous constants E, a, S, Y; K, K'; A, B, C describe the boson energies
and interactions. The derivation of these formulae is a straight-

forward application of group theory. The following group reduction
has been used
SU(b):::> SU(S) 8 U(i)

i) (3.4)
~ 0+(5)::J 0+(3)
ii) (3.5).
iii) (3.6)
The corresponding generators are:
(3.7)
i) U(S-)
ii) 5U(3) L(~ = (cl+d'(~) (3.8)
ex (2.) _ (elt
tt- S+5
t cd) (2)
f- -n (dtd)~}
2-
(3.9)
iii) O(~) (dfl);)) (d~!/)) (dls~$~)/2)
and these are seen to form subsets of the 36 generators g. of U(6).

~
Examples of the first two kinds of symmetry are shown in Figs. 2 and
3, respectively. By comparing the experimental levels with those
provided by the corresponding classification scheme, one can see
that the symmetries are not badly broken. In fact, they appear to
1050 F.IACHELLO
Exp Th Exp Th Exp Th Exp Th Exp Th

Y X Z f3 f::,.
E(keV) I I I I I
I 8+ 3482
6+ 3180
+3056
3000 5+ 2927/5 - -
+ 2162 + 2160 2+ 2287 -2+ 2285

3---3--
2000
o 1473 -0+ 1473
1000
2+ 658 _2+ 722
Fig. 2 An example of SU(5), vibrational, symmetry. The

experimental levels are denoted by Exp.; those pro-
vided by the classification scheme by Th. The sym-
bols Y, X, Z, S, 6 refer to the band classification
given in Ref.(4).
E Exp Th Exp Th Exp Th

(MeV)
1.5
4+ - 1298 - 1317 4 : - 1355 _ 1316
3+-1248 _ 1210
2+ - 1129- 1129 2-1154_1129
0+ - 1049 - 1049
1.0 8+ - 965 - 965
K=O K=2
(20,2)
6+ - 585 563
0.5
4+ - 288 268
2+ 89 80
0 0+ 0 0
K=O
(24,0)
Fig. 3 An example of the SU(3), rotational, symmetry.

The symbols (24,0), (20,2), K = 0, K = 2, refer
to the band classification given in Ref. (4).
be less broken than the corresponding dynamical symmetries in ele-

mentary particle physics. This is the more surprising when one
considers that nuclei are complex many-body systems and that the
nucleon-nucleon interaction has no simple invariance properties.
4. A GENERAL CLASSIFICATION SCHENE

F-SPIN
In the case in which one is close to the limiting situations,

as in the examples shown in Figs. 2 and 3, one can use the wave-
functions corresponding to the dynamical symmetries described above
as zeroth order and pecform small perturbation calculations around
it. However, many nuclei do not belong to any of the special
cases given in Section 3. It is especially for these nuclei that
the SU(6) classification scheme is very useful. In fact, since in
this approximation the SU(6) space forms the representation space
for any collective state, eigenstates and eigenvalues can be ob-
tained by diagonalizing the Hamiltonian H in this space, which ~s
a simple and straightforward numerical problem. An example of

the study of the transition from the SUeS) to the SU(3) symmetry
scheme is shown in Fig. 4. Here, the fact that the solutions are
confined within the group manifold, manifests itself in the unex-
pected behaviour of some of the states (for example, the excited
+
O2 state) as the classification scheme shifts from SUeS) to SU(3).
This state first moves down in energy, following closely the 4+
state but then turns upwards as the particle number is increased.
This is because in the SUes) limit this state belongs to the multi-
plet with nd = 2, while in the SU(3) limit it belongs to the repre-
sentation (A,W) = (2N-4, 2). The solution of the corresponding
geometrical problem is very cumbersome. From this point of view,
the SU(6) boson model provides a general and yet sim~le classifica-
tion scheme which can be used not only for nuclei which have spectra
typical of the limiting situations but also for all other nuclei
which are intermediate between them.
1052 F.IACHELLO
E
(MeV)
1.5
1.0
0.5
82 86 90 94 98
Neutron Number
Fig. 4 The transition from the SU(5) to the SU(3) clas-

sification scheme in the Sm-isotopes. The circles,
squares and triangles are the experimental ener-
gies, the lines are the results of the diagonali-
zation in the SU(6) space.
At this stage, it is worthwhile making a small digression.

The classification scheme of the previous section describes the
properties of a system of N identical bosons, corresponding in the
geometrical picture to one set of collective variables Ro,
a (~= +2, +1, 0, -1, -2). However, if we identify the bosons
~
as nucleon pairs we immediately recognize that one should actually
introduce two kinds of bosons, corresponding to proton and neutron
pairs, respectively6). The simplest way of discussing the coupled
system of proton and neutron bosons is that of introducing a for-
mal two-dimensional SU(2) space. This space is analogous, but
definitely not identical, to isospin and in order to avoid confu-
sion we have called it F-spin. The group structure of the problem
becomes now SU(6) ~ SU(2). It is relatively easy to see which
SU(6) representations are contained in the combined system. For

example, the reduction of the outer product [4J 0 [2] is
(4.1)
Since for identical bosons only the totally symmetric representa-

tion [6J is present, one sees that, by explicitly introducing the
proton-neutron degree of freedom, one obtains also states with
mixed symmetry ([5,lJ and [4,2J in the example above). The SU(6)
x SU(2) scheme thus provides an even more general classification
scheme in which states of mixed symmetry are described in addition
to the totally symmetric states. These states correspond, in the
geometrical description, to states in which protons and neutrons
do not move together, but perform oscillations and rotations in an
independent way.
5. ODD-A NUCLEI.
GRADED LIE ALGEBRAS Al~D SUPERSYMMETRIES.
The Lie algebras SU(6) (or SU(6) x SU(2)) provide a general

classification scheme for collective states in even-even nuclei and
give examples of dynamical symmetries in nuclear physics. The ques-
tion arises whether graded Lie algebras are useful in the context
of nuclear physics and whether or not nuclear physics provides ex-
amples of supersymmetries. The answer is again yes and I will dis-
. 7)
cuss here only a s~mple example Consider the problem of n iden-
tical nucleons in a single j-shell (for example, the Ca-isotopes).
Denoting the fermion creation and annihilation operators by a!,am
= (_)j-m a_m, one can introduce three operators:
(5.1)

1054 F.IACHELLO
+
These operators are scalars in ordinary space but form a vector S
in a fictitious space, called quasi-spin. Similarly one can intro-
duce the two operators
(2. j -4- of ) 'I.,. t
V, :.
8 I/l.
~
tV
"'I
(5.2)
(2j + 1 (Ii
VI
"'~i:
= 8 '/2-
tt
'V
These operators are tensors of rank j in ordinary space but form a

spinor. Ya is quasi-spin space. Using the definitions (5.1) and
(5.2) and the basic anticommutation relations of the operators
a!,am one can now show that
rL SA~ I Sb] == ~ _L
i C;~c. 5C a./b,c: 1,2,3 (5.3)
where S Sl iS 2 , S
Z
(5.4)
2.
where Ta are the Pauli matrices. Finally,
0.: 1,2,0. Ol,P>=.!.. _1-

) r 2' 2. (5.5)
where C is the charge conjugation matrix and contraction with re-

spect to ordinary angular momentum is always implied. The commuta-
tion relations (5.3), (5.4) and (5.5) are identical to those of the
graded GSU(2) algebra of Pais and Rittenberg 8 ). From this point of
view, a system of identical nucleons in a single j-shell forms an
interesting and non-trivial exampl~ of a graded algebra. One can
also enquire whether or not a classification in terms of the graded

algebra is useful. This will be the case if there are symmetries of
the Hamiltonian which are analogous to the dynamical symmetries for
normal Lie algebras. These supersymmetries will provide simple and
elegant solutions to the eigenvalue problem, give rise to selection
rules, etc., and will allow a general classification scheme in which
both fermion (n = odd) and boson (n = even) states are classified
within the same representation of the graded algebra. For the ex-
ample given above this appears to be the case. Thus, once more,
nuclear physics offers examples of supersymmetries which are at the
same time mathematically non-trivial and physically interesting.
The model of a single j-shell can only be applied to a small

number of nuclei (for example the Ca isotopes). In the more general
case, the fermion many-body problem is much more complicated. How-
ever, it may happen that, in analogy with the case of even-even nu-
clei, one may uncover supersymmetries which describe simultaneously
odd-A and even-even nuclei, by replacing the complicated many-body
system by a simpler system. It is clear that, when formulated in
our language, the algebraic structure of the problem is that of a
graded Lie algebra. In fact, in our language, even-even nuclei are
treated in terms of bosons while odd-A nuclei will be treated as a
system of bosons plus one fermion. Moreover, since the bosons are
fermion pairs, their creation and annihilation operators will not
commute with the fermion operators, much as in the example above.
Rather, they generate a graded Lie algebra. What is not at all
clear is whether or not the fermion-boson, fermion-fermion and boson-
boson interactions in the smaller space B have simple supersymmetry
character. If that is so, and work in this direction is in progress,
one will have a classification scheme, wherein both even-even and
odd-A nuclei can be simultaneously placed.
1056 F. IACHELLO
6. CONCLUSIONS
It is appropriate to conclude by summarizing some similari-

ties between the methods of theoretical analysis discussed here
and those used in elementary particle physics. In nuclear physics
we start from the complex many-body system, select some degrees
of freedom, those which we think play the fundamental role (the
boson), and study their symmetry properties. This has in turn
given us the full classification scheme of the collective states
in even-even nuclei and may give an even more ambitious classifi-
cation scheme in which both even-even and odd-A nuclei are simul-
taneously included. In elementary particle physics, one also
starts from some objects which are thought to playa fundamental
role (the quarks) and studies their symmetry properties. This
gives a classification scheme for elementary particles. However,
the generators of the dynamical symmetries in nuclear physics are
not elementary, nor have they elementary couplings with each other.
They are mathematical objects which replace the complicated and
usually unmanageable many-body system. In elementary particle
physics the generators of the dynamical symmetries, the quarks,
are thought to be elementary and to have elementary couplings with
each other. Thus, there appear to be some similarities and some
differences between the methods of analysis used here and those
used in elementary particle physics. w~ether these similarities
and differences have a deep meaning or rather they are accidental,
is not clear to me at the moment.
REFERENCES
1) E. Wigner, Phys. Rev. 51 (1937) 106.
2) J.P. Elliott, Proc. Roy. Soc. (London) A245 (1958) 128 and 562.
3) A. Arima and F. Iachello, Phys. Rev. Letters 35 (1975) 1069.
4) The vibrational, SU(5), limit is described in A. Arima and

F. Iachello, Ann. of Phys. 99 (1976) 253.
The rotational, SU(3), limit is described in A. Arima and
F. Iachello, KVI Preprint 105 (1977) to be published in
Ann. of Phys.
The triaxial, 0(6), limit will be described in a forthcoming
paper.
5) A. Bohr and B.R. Mottelson, "Nuclear Structure", Vol. II,

W.A. Benjamin, Reading, Hass. (1975).
6) A. Arima, I. Ohtsuka, F. Iachello and I. Talmi, Phys. Letters

66B (1977) 205.
7) F. Iachello, paper ~n preparation.
8) A. Pais and V. Rittenberg, J. Math. Phys. 16 (1975) 2062.

SYMMETRIES OF QUARKS AND LEPTONS*)
F. Gursey
Yale University, New Haven,
Conn. 06520, USA
1. INTRODUCTION - An Overview
The course is divided in four Chapters corresponding to three

lectures. After an introductory first chapter, the second deals
with the symmetries of strong interactions from a quark point of
view. We shall discuss the SU(6) x SO(3) symmetry of low lying
hadrons in relation to the field theory of quarks (chromodynamics).
The subject of the second lecture (Chapter III) is the unification
of the internal symmetries of quarks and leptons through exceptional
groups. We shall see how strong interactions, based on a gauge
theory of weak and electromagnetic interactions can be further uni-
fied in a gauge theory based on exceptional groups, E6 being the
best candidate. The third lecture (Chapter IV) is devoted to a
speculation concerning the origin of internal symmetries of quarks
and leptons. There are some unique exceptional geometries which
are connected with octonions and correspond to a finite number of
degrees of freedom that cannot be extended. This is the new
*) Research supported in part (Yale Report #COO-3075-211)

by the U.S. Dept. of Energy under Contract No. EY-76-C-02-3075.
1059
1060 F.GORSEY
Hilbert space that we want tentatively to identify with the Hilbert

space of internal symmetries carrying color and flavor quantum
numbers.
There has been a deep change in our conception of a fundamental

theory for elementary particles. In the last decade the elementary
particles were thought to be the classical leptons (e, ~ and their
neutrinos), and the hadrons (baryons and mesons). The photon
responsible for electromagnetic interactions interacted locally with
leptons, but non-locally with hadrons which have electromagnetic
form factors of the order of a fermi. Charged heavy intermediate
bosons were introduced to mediate weak interactions (instead of a
4-fermion interaction), but a renormalizable field theory allowing
higher order calculations could not be written for weak interactions.
No similar fundamental mechanism was available as a basis for strong
interactions which were thought to arise from hadrons exchanging
other hadrons following the thirty year old Yukawa mechanism.
The few clues about an underlying structure in hadron physics

came originally from their symmetry properties. First, there was
the 5U(3) flavor symmetryl). Hadrons fell nicely in eight and ten
dimensional representations of the group, but the fundamental
3-dimensional representation carrying fractional electric charge
and fractional baryonic number was missing. These were called
quarks 2 ). An approximate 5U(6) symmetry combining spin and 5U(3)
quantum numbers in a definite way accounted well for the low lying
hadron spectrum 3 ). This could be understood if the hadrons were
assumed to be some sort of composite structures of quarks, 3 quarks
for a baryon and a quark-anti quark pair for a meson 4 ). However
the mechanism for such a binding remained completely mysterious.
The crucial quark concept turned out to be very useful in weak and
electromagnetic interactions of hadrons since many of their
properties could be better understood if it was assumed that hadrons
SYMMETRIES OF QUARKS AND LEPTONS 1061
interacted with photons and.weak bosons through their quark

constituents.
The next step was taken by postulating a completely new hidden

symmetry based on the SU(3) group of color S ,G,7). This was necessary
a) to make the occurrence of symmetric SU(6) representations for low
lying baryons compatible with the fermion character of quarks, b) to
characterize observed hadrons by postulating that they should be
color singlets, c) to account for the rate of the decay 8) 1To+ 2 y
Meantime three momentous developments occurred. On the theo-

retical side the Weinberg9 ) -Salam10 ) theory was proposed as a
renormalizable gauge field theory spontaneously broken by the Higgs
mechanism. It was a Yang-Mills 12 ) type theory coupled to scalar
fields that succeeded in unifying weak and electromagnetic inter-
actions through a weak-electromagnetic local SU(2) x U(l) group.
It predicted a new type of weak interaction involving neutral cur-
rents. Those were discovered within a few years. The other experi-
mental discovery (scaling) concerned the point-like behavior
( partons 1 3 -14 ) ) of const1tuents
. .
w1th1n th
e nucIeon1n deep 1ne
. 1 ast1c
.
lepton-nucleon scattering. The theoretical discovery that provided

an explanation was the asymptotic freedom IS) of quarks interacting
through a gauge theory based on the local color grouplG) (a simple
non-abelian gauge group in general), assuming the group to be exact.
The force between the quarks (that were identified with the parton
constituents of hadrons) was mediated by the eight massless color
gluons that were the gauge bosons of a Yang-Mills theory based on
the color SU(3). This led to a completely new theory of strong
interactions based on a renormalizable field theory of unobserved
quarks having a long range interaction arising from the exchange of
unseen gluons. The new theory that has (apart from the nonlinearity
of the equations and the asymptotic freedom property) strong
1062 F. GURSEY
analogies with quantum electrodynamics (QED) was called

chromodynamics 17) (QeD).
In quark-field theory, quarks also have local interactions with
bosons associated with the weak-electromagnetic group SU(2) x U(l),
which, however, is spontaneously broken down to U(l). The weak
properties could only be properly accounted for at the expense of
introducing a fourth quark c carrying a new quantum number, charm 18).
The GIM 19 ) mechanism could then explain the absence of strangeness
changing neutral currents. The discovery of a new family of stable
hadrons 20) was well explained by the charmonium cc bound state 21 )
model and the predicted charmed mesons analogous to strange mesons
were also discovered 22 ). Then, the SU(3) corresponding to the
3 quarks u, d, s had to be enlarged to SU(4). Such groups
associated with the number of quarks became less fundamental than
the W-S. SU(2) x U(l) group or the color group SU(3).
The fundamental fermions are now the colorless leptons, en-

larged recently by the addition of the heavy lepton T- 23 ) and its
neutrino \! ~ to the list of classical leptons, the colored quarks
u, d, s, c and the new heavy b quark thought to be the constituent
of the newly discovered upsilon 24) levels (bb bound states).
The fundamental bosons are the photon, the color gluons

associated with the exact group U(l)el. x SU(3)col., the weak bosons
corresponding to the coset decomposition SU(2) x U(l)/U(l)el. of
the Weinberg-Salam group and possibly higher mass weak bosons and
superheavy bosons arising from the embedding of the SU(2) x U(l) x
c
SU(3) group in larger unifying local gauge groups. In addition
there may be some scalar massive Higgs particles.
The conventional strong interactions have taken the secondary

status of Van der Waals-like forces between color-saturated bound
states of quarks.
The new picture comes very near the realization of Einstein's

dream of a unified field theory. To achieve it, further unification

with the gravitational theory is needed. This can be partially done
within the framework of local supersymmetry, provided the internal
(color-flavor) group is small enough. To this date no unification
of the group SU(2) x U(l) x SU(3) with Einstein's theory and its
supergravity 25) extension has succeeded. When and if it comes about
we shall have to add the graviton to the list of fundamental bosons
and the gravitino (with spin 3/2) to that of basic fermions.
As we shall see, the origin of SU(2) x U(l) x SU(3)c group may

be sought in the properties of special finite Hilbert spaces which
also correlate the number of leptons and quarks. The SU(6) sym-
metry of hadrons involving hadronic spin arises from an approximate
spin independence of the long range force between quarks in chromo-
dynamics. This is a dynamical problem which is poorly understood
since it depends on a non-perturbative treatment of QeD. The spin-
dependent forces that break SU(6) symmetry are better understood as
they come from gluon exchanges that can be calculated in perturb-
ation theory. Anyway, the hadronic SU(6) symmetry which is part
of an SU(8) symmetry (with charm) or SU(lO) symmetry (with the in-
clusion of the b quark) cannot be fundamental. The hopes raised by
the discovery of supersymmetry26) between fermions and bosons for a
better understanding of symmetries 27 ) involving spin have faded
away. The fundamental role of supersymmetry seems to be in super-
gravity and not in the phenomenology of hadronic mass levels and
interactions.
The fundamental questions have now shifted to the understanding

of the underlying quark substructure of hadrons. We would like to
understand the emergence of the hierarchy of interactions, the mass
spectrum of quarks and leptons and the wider mass spread of the
fundamental gauge bosons through the spontaneous breaking of a
1064 F.GURSEY
local symmetry based on a unifying group or supergroup associated

with a gauge field theory.
The concept of a field theory that unifies both the various

kinds of interactions and the great variety of symmetries of funda-
mental processes is philosophically and aesthetically more satis-
fying than the old division of labor, concepts and methods in the
main areas of physics; geometry for gravitation, field theory for
electromagnetism, effective field theory for weak interactions and
S-matrix methods (based on analyticity and unitarity) for strong
interactions. The only trouble with the new natural philosophy is
that it deals largely with an unobservable substructure (quarks,
color quantum numbers, superheavy gauge bosons, hidden massless
gluons, essentially undetectable gravitinos, the 8-dimensional
superspace) which makes the physicists rather uncomfortable and
reminds them of other imponderables of the past (the caloric fluid,
the aether) ridiculed by later generations.
The confrontation of the new theory with experiment is by its

very nature indirect, in contrast to the foundations of twentieth
century physics (special relativity and quantum mechanics) that had
their roots in the principle of dealing only with observables. The
S-matrix theory of strong interactions came from an extension of
the principle and dealt with observable scattering transition prob-
abilities. We are now back to a view more akin to the one held by
Einstein in his old age, spinning out geometrical structures far
removed from the space-time geometry based on the equivalence prin-
ciple, trying to unify gravity with other fields that seemed to
have non-geometrical properties. It is ironical that Heisenberg
also in his old age rejected his positivist S-matrix approach in
favor of an underlying (non-linear but non-renormalizable) field
theory for the unification of electromagnetism and strong inter-
actions.
In defense of the new philosophy it might be said that other

past substructures that were regarded as imponderable or too abstract
by contemporary physicists did prevail in the end and were accepted
as physical reality. Atoms, fields (both electromagnetic and gravi-
tational), antimatter, the elusive neutrino, the particle-wave dual
nature of matter and radiation, Hilbert space, the complex wave
function, etc., might provide a few examples.
In these lectures we shall start from the experimentally

accessible, phenomenological SU(6) symmetry 3) and chiral symmetrl8)
of strong interactions, try to understand them as a reflection of
some properties of the quark substructure. Then, we shall note the
similarities between the synlffietries and the local interaction
properties of quarks and leptons leading to an extended notion of
quark-lepton symmetry which strongly suggests that both quarks and
leptons are members of the same multiplets 29 ). This leads directly
to the notion of a unified gauge field theory for electromagnetic,
weak and strong interactions that is a generalization of electro-
magnetism based on a local U(l) gauge group. An example of such a
gauge field theory based on E6 will be discussed 30 ). It extends
previously suggested unified field theories based on SU(5)31) ,
SO(10)32) , SU(3) x SU(3)C 33) or SU(6)34) Some general problems
of this type of unification (quasi-stable proton, superheavy mass
scale, big renormalization effects, "unnatural" features of the
resulting hierarchy and mixings between the particles) will appear
as blemishes of this otherwise attractive unification. Possible
alternatives will be briefly mentioned. Finally, we shall move to
a still more abstract plane, in fact the octonionic planes that may
provide the geometrical foundation for the existence of internal
symmetries like color and flavor. These octonionic geometries
allow us to construct new finite Hilbert spaces with unique pro-
perties. The non-Desargues'ian geometric property makes them
1066 F. GURSEY
non-embeddable in higher spaces, hence essentially finite. It also

leads to peculiarities in the superposition principle in the color
sector of the Hilbert space. This theory of the charge space, if
correct, suggests a new geometric picture for the substructure of
the material world, that of an octonionic geometry attached at each
point of Einstein's Riemannian manifold for space time with local
symmetry that leaves the properties of charge space invariant.
The text is an updated version of the original 1977 lectures

and contains references not available at the time.
II. HADRONIC SYMMETRIES and the QUARK SUBSTRUCTURE
1. g~~~!~~!!~~_~!_!~!~!~~!_~~~_~E~~:!!~~_~~~~!!!~~
This chapter is partly relevant to the question put forward by

Professor zichichi in his introductory lecture: "Can Internal and
Space-time symmetries be unified?". It will deal briefly with
hadronic symmetries that unify approximately internal symmetries
with spin (SU(6) type symmetries) and chirality (chiral SU(2) x
SU(2) type symmetries). Since spin and chirality are space-time
concepts, the groups in question lead to symmetries that unify space-
time and internal properties. Both approximate groups are very good
phenomenologically, in fact as good as isospin for SU(2) x SU(2) or
the flavor SU(3) for SU(6). Is there anything fundamental in the
validity of these groups?
From the point of view of modern gauge field theory connected

with the quark substructure these groups are thought to arise as
dynamical symmetries related to the spin independence of the long
range part of the quark-quark potential in the case of SU(6) or
accidental symmetries due to the smallness of the masses of the up
(u) and down (d) quarks compared with masses of the other quarks
(s, c, b, ... ) in the case of the chiral SU(2) x SU(2). Hence there
seems to be nothing fundamental about these flavor groups or their

combination with spin and chirality. Their validity depends on
approximate mass degeneracies between the quarks and ill understood
dynamical properties of gauge theories. Hadronic isospin is good
because u and d are approximately mass degenerate, chiral isospin
is valid because both masses are small, SU(3) is broken because the
strange quark s has a higher mass than u and d, also leading to a
large breaking of chiral SU(3) x SU(3). SU(4) is broken even more
because the charmed quark is so much heavier, and the chiral SU(4)
x SU(4) is no longer useful. The same applies even more strongly
to the SU(5) connected with the b-quark. As a result, the problem
of understanding the success of some of these phenomenological groups
unifying internal with external symmetries is replaced with the
problem of understanding the quark mass spectrum and the properties
of quark-quark interactions. Quarks must behave like free particles
at short separations so that their spin can be conserved and behave
like an internal quantum number to be unified with the other quark
degrees of freedom. This part is well understood through asymptotic
15)
freedom At larger separations gluon exchange takes place
resulting in a Coulomb-like potential and spin dependent effects
that break the spin independence 35) The symmetry breaking terms
in SU(6) can be derived from the perturbative calculation of spin
dependent forces in QCD. However both the spin independence of the
long range confining potential and the quark mass spectrum depend
on the non-perturbative solutions of Quantum Chromodynamics and
Quantum Flavor dynamics. Since we are far from knowing these
answers we shall replace the fundamental theory with plausible
effective field theories for hadrons in order to have a quantitative
framework for the phenomenological hadronic symmetries.
Are there other, more fundamental ways of combining internal

and external symmetries?
1068 F. GURSEY
a - Gauge Theories provide one kind of combination if internal

symmetry groups are regarded as locally valid. The parameters of
the internal symmetry group are arbitrary functions of x. Certain
functions which tell us how the directions of internal space are
oriented from point to point in real space time will give correlat-
ions between internal symmetries (color, isospin, weak isospin) and
external symmetries (like spin direction at each point) that may
result in a lower energy configuration or a minimum of the action
in Euclidean space-time that can be interpreted as tunnelling
between various vacua. This is exactly what happens in monopole
and instanton solutions in gauge field theories. Since the subject
is amply treated ~n Prof. Coleman's lectures 36 ) I shall not dwell
upon this aspect of the fusion of space-time and internal properties.
Clearly what is going on for such solutions is more fundamental than
the approximate fusion stemming from the lightness of some quark
masses.
b - Another more intimate blending of internal and space-time

26 27)
symmetries occurs in supersymmetry ~n general' and extended
37)
supergravity in particular. SU(6) type symmetries involve sym-
metries between bosons of different spin (0 and 1) or fermions of
different spin (1/2 and 3/2). In supersymmetry we have a more
general symmetry between bosons (integer spin) and fermions (half
odd-integer spin). The symmetry also survives a local interaction,
a feature not present in SU(6) symmetry where the interaction term
is invariant under a non-local group (SU (6)). Thus, super-
w
symmetry has a more fundamental field theoretical character. But
it seems to have nothing to do with the approximate hadronic sym-
. 25) .
metries involving spin. In pure supergrav~ty ~t leads to a
symmetry between the spin 2 graviton and the spin 3/2 gravitino.
In extended supergravity with U(l) as internal group the symmetry
is between the graviton, the photon (spin 1) and spin 3/2 gravitinos.
SO(3) also brings in scalar (spin 0) fields in the multiplet. The

internal group can in principle be extended to 0(8) which, however
is not large enough to accommodate the minimal flavor-color group
SU(2) x U(l) x SU(3)c. Efforts to break the 0(8) barrier in super-
39 40)
gravity or conformal supergravity are under way' For the
moment we can tentatively state that supersymmetry does achieve a
field theoretical unification of internal and space-time symmetries,
but this unification is not relevant to the approximate phenomeno-
logical unification of spin independence and flavor groups in
hadronic physics. The absence of lectures on supersymmetry in the
present course is unfortunate from a fundamental point of view, but
justified on the grounds of relevance to the present high energy
physics. Supersymmetry and supergravity will undoubtedly be fully
treated in the future Erice courses on particle physics.
2. ~~~E~E~~l_~~E~E~EiY~_~~_Y~~!Ei~~_~i~~~g_Ei~_~~_l~!~E~~l
g~~~E~!!.l_~~~E~
I shall be brief ~n my review of the SU(6) hadronic symmetry

since it has been the object of numerous Erice Lectures in the past
41
between 1964 and 1976 ) Here are some brief historical notes.
The earliest consideration in 1937 of a symmetry between states
of different spin and isospin is due to Wigner42 ). The motivation
came from nuclear physics. The symmetry was given a mathematical
structure through Wigner's construction of the generators of the
group SU(4) from spin operators 1/2; and isospin operators 1/2 ;
acting on a 4-dimensional space of an isodoublet with spin up and
spin down. This fundamental representation corresponded to the
4 nuclear states pt, p+, nt, n+. The fifteen generators of the Lie
algebra, in the 4 x 4 representation, are
(2.2.1)
1070 F. GURSEY
Light nuclei were associated with higher representations of

SU(4). The 4 nucleon states are (apart from the small electro-
magnetic mass difference between p and n) degenerate. When we
consider nucleon bound states corresponding to a given baryon number
there will be mass degeneracy (apart from Coulomb effects) in a
SU(4) multiplet if the nucleon-nucleon potential is spin independent
(of the Majorana form). The symn~try will only be broken and the
mass levels split by spin-dependent parts of the potential. This
works well with light nuclei. In heavy nuclei the Coulomb effects
are too large to make the symmetry a useful one. As shown by
Radicati and Franzini43) applications still exist in that case.
The really useful symmetries in heavy nuclei, which again put levels
with different spin in the same multiplet are the ones proposed by
Iachell0 44 ) and his collaborators.
The next step was taken independently from Wigner by Moeller

and Rosenfeld 45 ), Stueckelberg 46 ), Rarita and Schwinger47),
48) 49)
Schwinger , and Jauch during the war years. Moeller and
Rosenfeld rediscovered (but did not recognize) the SU(4) symmetry
in a field theoretical context by assuming that nucleon forces
arose from the Yukawa exchange of vector and pseudoscalar mesons
with same vector and pseudovector coupling to nucleons and with a
mass degeneracy. Such a "mixed" theory of nuclear forces yields a
spin independent potential, hence an SU(4) symmetry of nuclear
forces. The correct number of mesons that had to be exchanged to
achieve such symmetry was given by Stueckelberg. They correspond
to negative parity mesons (vectors and pseudoscalars) associated
with the generators (2.2.1) of SU(4). If we use modern notations
for the mesons and show their correspondence with the SU(4)
generators we have,
(2.2.2)
(2.2.3)
namely, the wand p vector mesons and the pions that form the
15 dimensional multiplet of the adjoint representation of SU(4).
The resulting spin independent potential was applied to N-N scatter-
ing and the deuteron by these authors who were unaware of Wigner's
work and failed to recognize the group structure of their theory.
Again, without the use of group theoretic language, Rarita and
Schwinger introduced a symmetry breaking in the theory by keeping
the vertex symmetry of the couplings to obtain a well-behaved
potential for the Schrodinger equation but by breaking the symmetry
through a mass difference in the 7T and p mesons. They were thus
able to explain the electric quadrupole moment of the deuteron
49) .
which vanishes in the exact SU(4) symmetry limit. Jauch , us ~ng
the same theory showed that the best fit to experiment is obtained
if the p meson is a few times heavier than the pion.
Unfortunately, all this beautiful work on an implicit SU(4)

theory and the resulting predictions of the w, 7T and p mesons with
the correct quantum numbers and the correct mass splittings were
totally forgotten after the war with the discovery of the pion and
the advent of the pion theory of nuclear forces which did not
involve the wand p mesons. Their discovery came more than a
decade later. Another reason for the hostility towards a charged
vector meson can be sought in the renormalizability for the 7T-N-N
interaction which does not hold for the hypothetical p -N-N coupling.
In those days nucleons and pions were associated with fundamental
fields interacting locally, and non-renormalizability was a sin,
when quantum electrodynamics held the center of the stage as the
1072 F. GURSEY
most accurate theory devised by man. With our present-day wisdom,

we know that the renormalizability criterion in strong interactions
should be applied to the interaction of quarks with gluons. The
nucleon meson interaction can at best be described by an effective
Lagrangian that requires no renormalizability and consequently
cannot single out the pion interactions at the expense of vector
meson interactions. Instead, the meson effective interactions are
smoothed out by their form factors that introduce a natural cut off,
so that a nonsingular potential that allows the Schroedinger equat-
ion to be solved emerges from the exchange of pseudoscalar and
vector mesons. This principle of maximum smoothness for the potent-
ial is exactly the one used by Moeller, Rosenfeld and Schwinger.
After the wand p mesons were well established, and the SU(3)
flavor symmetry gained acceptance through the completion of the
pseudoscalar and vector octets, it was time to redo the work of
the early forties on the baryon-baryon potential by allowing them
to exchange vector and pseudoscalar octets, replacing isospin by
SU(3) and upgrading Wigner's SU(4) symmetry to SU(6) with the
35 generators
( n 1,2,3; a 1, . ,8), (2.2.4)
acting on a 6-dimensional representation of SU(6) corresponding to

the (2,3) representation of its maximal subgroup SU(2)spin x SU(3).
This (6) of SU(6) is a spinor in space-time and a triplet in SU(3)
space. We recognize it as the quark representation q which takes
over the role of the nucleon in Wigner's SU(4). The low lying 0
and 1 mesons fit in the adjoint representation contained in
6 x 6, hence it can be symbolically designated by qq, while the low
lying baryons, the (1/2)+ octet and the (3/2)+ decuplet correspond
to the 56 dimensional representation contained in the symmetrical
part of 6 x 6 x 6 symbolized by qqq. Similarly the antibaryons are

1n 56 (qqq). This is the first indication of a quark substructure
of baryons. Mesons and baryons can be regarded in some sense as
quark-antiquark and 3-quark bound states respectively. If the
quark binding force is approximately spin independent and SU(3)
invariant, the meson and baryon multiplets will show approximate
mass degeneracy. The breaking of the spin is due to spin dependent
and strangeness and isospin dependent terms in the interaction.
Again the irony is that these developments occurred without the
knowledge of the mixed meson theories of the forties.
The main difficulty in accepting the quark substructure was

the symmetric character of the (56) baryon representation. If (qqq)
are in an S state as suggested by the spins of the baryons, with q
having spin 1/2, the baryons should be in the antisymmetrical (20)
representation. The paradox of the symmetric representation was
solved by Greenberg 5 ) who introduced parastatistics, essentially
equivalent to the tripling of quark states. This was recognized
6) ..
by Han and Nambu who, however ass1gned 1nteger charges to quarks.
The tripling of fractionally charged quark states, more akin 1n
spirit to the original Greenberg picture, was proposed by
Gell-Mann7 ) who also gave the name color to the new degrees of
freedom associated with a new SU(3) that turned out to be more
fundamental than the flavor SU(3).
Another problem with SU(6) symmetry was the formulation of

an exact SU(6) limit in relativistic field theory. Although the
symmetry emerged in the non-relativistic limit, it did not exist
in the relativistic case except for free particles. This was
. 50) . 51) , . 52) .
p01nted out by Coleman ,We1nberg and 0 Raeferta1gh Th1s
is in contrast with supersymmetry in which the symmetry between
fields of different spin which include fermions survives in the
relativistic local interaction limit.
1074 F.GORSEY
If we limit ourselves to SU(6), the vertex symmetry for the

baryon meson coupling involves a rearrangement of the meson fields
and derivatives 1n the (35) multiplet. The resulting symmetry,
53)
first exhibited by Stech was rediscovered and written explicitly
54)
in this case by Lipkin and Meshkov and called SU(6) symmetry.
w
Both the static SU(6) and the vertex SU(6) are subgroups of SU(6)
w
x SU(6) that in the extreme relativistic limit admits the chiral
SU(3) x SU(3) as another subgroup.
Starting from the charge algebra of chiral SU(3) x SU(3) and

relating it to the current algebra (charge densities) of vector
and axial vector currents, the chiral algebra can be enlarged to
SU(6) x SU(6). This provided an alternative way of formulating the
commutation relation of SU(6) which should be valid even when the
symmetry is broken55 )
It gradually became clear that the best way of extracting

SU(6) symmetry and its breaking was through a field theory of
quarks. If an interaction potential is postulated among quarks,
the emerging bound state structures for mesons and baryons are
already fairly realistic as shown by Dalitz 4) and Morpurgo 4~ With
the advent of chromodynamics 8,9)based on a local gauge theory of
color SU(3) we have a framework for the possibility of calculating
the quark quark interaction or the potential when the potential
approximation is permissible. The general features of the short
35)
range part were shown by Glashow et al., to give the right kind
of spin dependence for the breaking of SU(6). This part comes from
gluon exchange and can be calculated in perturbation theory. The
long-range interaction was shown by Wilson 56) to be calculable by
a non-perturbative lattice approximation that resulted in a linear
confining potential, also 1n agreement with the string picture of
dual theories. This part of the potential is essentially spin
independent, leading to the spin independence of the mass spectrum

of bound states.
The success of free quark field theory and the connection

between current and constituent quarks given by free field theory
. ) 57) .
(the Helosh transformat~on could also be understood ~n QCD.
15)
Because of the asymptotic freedom property ,quarks behave like
free fields (partons) at short separation. This also explains the
success of current algebra based on free quarks.
The corrections to the free quark SU(6) come chiefly from the
distortion of free particle wave functions due to the confining
potential. These effects can be calculated in the bag approx-
. .
~mat~on
58,59) and resu It~n b et t er agreement w~t
. h exper~ment.
.
The last major development associated with the phenomenological

validity of SU(6) came from a better understanding of the breaking
of the SU(6) vertex symmetry for decay processes. The SU(6)
w
algebra is unchanged if we replace the spin operators d by u~ut
where U is a unitary matrix depending on the position x or the
momentum p. Unitary transformations of the Hamiltonian ~n an
effective Lagrangian theory will induce such transformations on the
. the Fo ldy-Wouthuysen trans f
spin operators. An example ~s
ormat~on
60)
. . 61) . 57). .
as po~nted out earl~er The Melosh transformat~on ~s a s~milar
example. This modified form of the vertex symmetry turned out to

account very well for the decays of the baryons and their orbital
excitations. Applications were also given to photoproduction
processes. The work was mainly developed by Gilman, Kugler and
62) 63) 64)
Meshkov ,Hey, Rosner and Weyers ,Close, Osborn and Thompson
After the discovery of charm, interest shifted to chromo-

dynamics and the SU(6) symmetry as well as its SU(8) version includ-
65)
ing charm were re-evaluated from the QCD point of v~ew.
As mentioned before, attempts to give back the phenomenological

1076 F. GURSEY
SU(6) a more fundamental status were made27 ) following the discovery

of supersymmetry. To date this approach has not yielded any real-
istic model that exhibits a hadronic SU(6) symmetry. The emergence
of SU(6) or SU(8) together with its breaking as an approximate
dynamical symmetry in QCD seems more probable at present. We may
give as an example a suggestion by T.D. Lee and R. Friedberg66 ) for
extracting an effective spin independent scalar potential from QeD.
3. ~~~~~~~~!~gi~!_Y~!i~i~~_~!_~~_~EE!~!i~~~~_~gi2_~_~Qi~2orbital
:_Q~_~~_g~~~!2~~~~_!~i2_~~~!ll
Let us briefly review the experimental evidence for a broken

SU(6) x SO(3) where the SO(3) refers to orbital excitations L of
the quark bound state
. 67)
a) SU(6) x SO(3) symmetry of the mult1plet structure
qq meson multiplets that are experimentally well established:
l(L=O); 35(L=0); 35(L=l).
qqq baryon multiplets that are experimentally well established:
56(L=0); 70(L=l); 56(L=2).
There is also the new possibility of the existence of
qqqq exotic mesons that would fit into 405(L=0) or 189 (L:O)
b) Decays between multiplets.

Breaking of the symmetry causes mixings like that of
56(L=0) with 70(L=l). This allows us to correlate a
number of amplitudes. The results are in impressive agreement
. 62,63,64)
w1th exper1ment
c) Various regularities related to SU(6) symmetry.

Exact SU(6) gives relations which are in most cases good
to within 30%. Confinement corrections always improve the
agreement with experiment in the right direction.
Examples are:
The ratio of the proton and neutron magnetic moments ~p/~N is
-3/2 from SU(6), the experimental value being -1.46. Related
results are the vanishing of the charge form factor of the neutron
and the constancy of the magnetic and electric form factor ratio
GM(t)/GE(t) where t is the invariant momentum transfer.
The F/D ratio associated with the symmetric and antisymmetric

SU(3) couplings of the pseudoscalar octet: F/D = 2/3(1+K tan)
where = 0 in the exact SU(6) limit, being the mixing angle
determined by decay processes. For 0 we have a = F/(F+D) = 0.40.
69)
The experimental value is 0.39 0.09, so that the corrections
are very small.
The magnetic dipole transition for ~ - N coupling for the
vertex (~ - N - y) is found to be
~*/~ 2/312 ~ (2.3.1)

p cos
This is off from the experimental value by 30% if and K

are neglected. The ~ - N- 'IT width which is 78 MeV comes out to
be 120 MeV in the exact SU(6) limit. Thus corrections are import-
ant for ~ processes.
For the gA/gV ratio the exact symmetry limit gives 5/3 which
is too high. However confinement corrections bring it down to
59
values between 1.20 and 1.30 in agreement with experiment ).
From this brief survey we can conclude that the SU(6) structure
is undoubtedly present in hadron physics. The question is:
how far is it understood and how fundamental is it?
Understanding is always done relative to some concepts and

theories that take their place within the paradigm of the era.
In the 19th century phenomena were understood if they could be
1078 F. GORSEY
described by Newtonian l1echanics. In the 1930's problems were

solved by determining particle wave functions that satisfied a
Schrodinger equation with a given potential. In the fifties under-
standing meant derivation from field theory through perturbative
methods. In the sixties physicists, nurtured in analyticity and
unitarity, tried to understand symmetries like SU(3) and SU(6) in
a bootstrap framework. The paradigm of the seventies is once again
quantum field theory in the special version of gauge theories.
Hence, today, understanding SU(6) means relating it to supersym-
metric field theories, to QeD, or to effective field theories that
can be regarded in some sense as good approximations to the under-
lying local gauge theory. In the following we shall review some
of these approximations.
a- Free quarks, leading to the conservation of the Foldy-

Wouthuysen spin and separate conservation of quarks and antiquarks.
This is a short distance approximation to QeD.
b- Heavy quarks, leading to the nonrelativistic limit of QeD

with a potential the short range part of which can be calculated
in perturbation theory.
c- Lattice gauge theory approximation for calculating the long

range confining part of the potential.
d- Once an approximation for the quark quark potential is

obtained we can calculate bound states using the Schrodinger theory
(nonrelativistic case) as a limit to the Bethe-Salpeter equation.
e- The qq bound states (mesons) can be described approximately

by local fields and their contribution to the force between baryons
(qqq bound states also approximated by local fields) can be calcul-
ated in an effective Lagrangian theory for baryons and mesons.
This gives a modern version of the work of the forties on an impli-
cit SU(4) symmetry.
f- The vertex symmetry SU(6) and decay processes can again

w
be studied by the effective Lagrangian method.
g- Since effective Lagrangians are widely used in hadron

physics in connection with chiral symmetry we can also discuss the
combination of these two kinds of symmetry involving space-time
properties by the same method.
In the following \ole shall sketch some of these approximations.
First consider the simplified case of colored quarks with no

flavor, interacting through the exchange of color gluons and obey-
ing the Dirac equation
i
1l11J!, (i 1,2,3), (2.4.1)
where m is the quark mass, i the color index and D~ the covariant
derivative.
~ a a
D ' \ + zgA BI-l' (a 1, . ,8), (2.4.2)
).l
where g is the quark-gluon coupling constant and theA's are the

eight 3x3 Gell-Mann matrices associated with SU(3). When we con-
sider a quark and an antiquark localized at very short distance
separation, then, from asymptotic freedom 69 )we know that the
renormalized strong fine structure constant g2/4TI
r
= a s tends to
zero at high momentum transfers (short spatial separation). In
such a situation the covariant derivative D~ can be replaced by
d and in the zeroth approximation we can treat the quark inside
).l
the hadron as obeying the free Dirac equation for small x, the
origin being the center of mass of the quark system in the hadron.
1080 F. GURSEY
Then we can write
(2.4.3)
where
A
b :: (2.4.4)
(2.4.5)
if Y5 ~s diagonal, (2.4.6)
and
if Y4 ~s diagonal. (2.4.7)
} ,bt (s = 1,2) are respectively creation operators for

p,sz p,sz z
particle and antiparticle Wigner states for the Poincare group
corresponding to momentum p and spin component sz' i.e.
1m, s
-7
,k.
2'
p, S >
Z (2.4.8)
with
l
m (2.4.9)
The exponential can be evaluated as
(2.4.10)
We also have the useful identity
-+--l'.!
{ x;:; r(.)} -t+~lf.
ex.p "2 u~~) f\ P
I
-2-
_ ,/PO(tt1+i.~'e)21+~1t
- V~ f.. 2' (2.4.11)
The projection operators 1/2 (1~4) project the large (+) or small
(-) components of the free wave function. To simplify the formulae
we have also omitted the color index.
We have the following conserved quantities
a) Quark Number
(2.4.12)
where
.
J'~ ;:::
-L\((Jl
Ip U r ) Col = 1)2,3 ) . (2.4.13)
J p )
This is the number of quarks minus the number of antiquarks.
b) Charges
Putting back the flavor index r in the case of 3 quarks
(2.4.14)
1082 F. GORSEY
c) Energy Momentum
(2.4.15)
The Hamiltonian pO is given by:
(2.4.16)
d) Conserved Spin Operators

.70) 60)
We now make the Tan1 -Foldy-Wouthuysen transformation
which is the same as the Wigner transformation on particle states.
Let us rewrite the Dirac equation in the form
(2.4.17)
(ia
n
= pn on a momentum eigenstate).
The covariant states are obtained by applying on the vacuum

the operator ~c = Y2~ *, the charge conjugate operator obtained by
exchanging a's and b's.
(2.4.18)
where we have used

(2.4.19)
and the identity (2.4.11). To obtain the covariant state from the
Wigner state we must then apply a non-local operator on
I p > exp(i p.x). We find
'P'lo>:= W(-i. :~1).f ~fe~P'\4~p)ID (2.4.20)
where
W(-I,. V)
........
- = r-m-+ IS( \7l*

- M
1.
-\J /2
Ji '
I
(2.4.21)
so that the operator
(2.4.22)
c
creates Wigner antiquark states and ~ (x) creates quark states.
The non-local operator W(-iV) is a unitary operator that connects
covariant states with Wigner states.
We can now transform the Hamiltonian with W. Transforming

(2.4.17) we find
H'.: WHW-'= '~4 E(O ::. ~~Jm1.+ Ip/2. = ~)m~v2)~ (2.4.23)
Thus, the transformation diagonalizes H in a representation for

which ~4 is diagonal. The expansion of (2.4.23) in 2 gives m-2v
the Schrodinger operator for the upper and lower components
separately, plus relativistic correction terms. Note that HI,
unlike H, commutes with the spin operators crn and with Y4cr n
(2.4.24a)
or
(2.4.24b)
Hence, the non-local objects
(2.4.25)
are constants of the motion. These are the non-covariant spin

operators that are conserved for free quarks and approximately
conserved in QeD. Thus the conserved generators of the relativi-
stic but non-covariant SU(6) x SU(6) are
(2.4.26)
1084 F. GURSEY
If we work with the set (2.2.4) we must use the transformed

Hamiltonian H'.
We now turn to another transformation of the Hamiltonian that

singles out a direction and therefore is valid on the light cone.
-2 2
For extremely relativistic quarks the expansion of H in m V or
the velocity square is not a good one. In this case we can remove
transverse components of p, say p~ (PI and P2 if the direction
singled out is the third axis) by a unitary transformation. This
is the Cini-Touschek 1 ) transformation rediscovered by Melosh 57).
It l.S given by
W := (2.4.27)
J.
where
..... -
~.L' p~ .: ~, P. +~.! F:2. '
./2.
F'.l. = V P. + P:4 ,
2
(2.4.28)
f" .= - i 6n (2.4.29)
Then the transformed Hamiltonian takes the form
Note that HI! no longer commutes with n and Y40 n but only
with the subset
(2.4.31)
that also generates an SU(2) group.
also generate rotations and the generators of the SU(6~ group
...!.. f5' /I (2.4.33)

2.. n ,
commute with the original Hamiltonian H.

53)
The operators (2.4.31) are the Stech spin operators. They
were rediscovered and called W-spin operators by Lipkin and
Meskov 54 ). We have seen that they are not conserved quantities,
but their Cini-Touschek transforms 0" are conserved in the free
n 57)
quark limit. This property was discovered by Melosh
The importance of the spin operators 0" is that they also

n
occur in the symmetry of the vertex in which the singled out
direction is given by the momentum transfer.
The meaning of Y4 can be elucidated by exhibiting the trans-

formation properties of the conserved quantities 72) We distinguish
between C-odd and C-even conserved covariant spin tensors. The
C-even tensor is given by
(2.4.34)
where
- -
Ht' yA = -t ~ ,
'f cr')A\> ()~ ~ = -~ E)L))A~ If 6, ~f '1>+ add. terms
(2.4.35)
is conserved:
"0 A H)LV" = 0, (2.4.36)
so that the n~v are time independent. These are related to the
Wigner spin as
-I
'vJ SG OL W = 0 ~
0
-2IE ..
lJ K
W-I .Jc.
n. u W = 52:.:.-
J '" l 2.
'y(t a.o.
a.. P L P
-to bt d. b)o3
r L
dp
p (2.4.37)
1086 F. GURSEY
Hence they can be interpreted as quark spin plus anti quark spin and
are associated wi th the d ~
1
Note that~. generate SU(2) but not the covariant quantities

1
The C-odd conserved covariant tensor is
(2.4.38)
where
and
(2.4.40)
Making the unitary W transformation we find the non-covariant

objects
Jl:::. W-'n
t.
5i W ~ ~f( ~t cr. a. - bt G, b) io
pl.,. pl.~ I'
(2.4.41)
-1
W 52!.;o W-O. (2.4.42)
Clearly these conserved quantities are the quark spin minus

the antiquark spin and correspond to X4~i' 1/2 (n. ~!) generate
1 1
the SU(2) x SU(2) group of quark and antiquark spin. They corres-
pond to 1/2 (1 a4)~i' These are sometimes called the f spin and
the f spin.
When we do the Cini-Touschek-Melosh transformation on J2~~and

Sl:,JA-' the C-odd operators .l2.i. Sl.; and the C-even ~ COIImlute wi th the
Hami 1 tonian H".
If" now we let P3 -+ 00 then
and the C-even subgroup of SU(6) becomes the chiral SU(3) x SU(3)
group whose generators COIImlute with massless Dirac Hamiltonian.
so that
li tn [SLJ(3)
~~ "~DO
To sum uP. the free Dirac Hamiltonian in the Foldy-Wouthuysen

form admits a SU(2N) x SU(2N)- sYIImletry where N corresponds to
q q
the dimension of the internal space and q. q refer respectively to
quarks and antiquarks. If a longitudinal direction is singled out
by an additional term in the Lagrangian. such as an interaction
term. then the Dirac Hamiltonian can be transformed by a Cini-
Touschek transformation relative to the longitudinal direction
and the Hamiltonian has a SU(2N) invariance associated with the
w
subgroup of SU(2N) x SU(2N)_ involving the C-odd transverse spin
q q
and the C-even longitudinal spin.
Consider the coupling of the fermion 1/J to a boson </>. If the

virtual boson with momentum q is emitted by the fermion 1/J (p) in
momentum space so that the fermion wave function changes from
1/J(p) to 1/J(p'). then p - p' = q is the momentum transfer. 1/J(p) and
1/J(p') may refer to states with different internal quantum numbers
such as u and d quarks. In the approximation that these states are
degenerate in mass the momentum transfer q is space-like and
singles out a space direction. Thus a Yukawa interaction term will
1088 F. GURSEY
involve a direction in momentum space (the q direction) which can

be taken as the direction singled out by the Cini-Touschek
transformation.
Then the interacting Hamiltonian can have a non-local

(q-dependent) SU(6) symmetry, provided the free boson Lagrangian
w
also exhibits an SU(6) x SU(6). symmetry generated by C-even and
q q
C-odd spin operators. Hence we need two kinds of bosons to achieve
vertex symmetry: C-even bosons cp and C-odd bosons wfJ. Those which
have the same parity can be grouped in the same multiplet. Negative
parity bosons will transform like the S-states of the (q q) system,
PC
so that we can choose multiplets of J bosons which transform
-+
like ljiYslji (0 ) and lji y lji (1 ).
fJ
Let us write such a Yukawa coupling in the case of no internal

symmetry
, (2.5.1)
(2.5.2)
(2.5.3)
where.t (lji), (p) and.,t' (WlI) are respectively free Lagrangians for
o 0 0 ~
the fermion, the pseudoscalar and the vector bosons, wi th.t (w )

o fJ
chosen so that the equations of motion also yield the subsidiary
condition
(2.5.4)
essential for associating object, and.t.

wfJ with a 1
is the
1nt
simplest non-derivative Yukawa coupling for the fermion-boson
system.
\ole could now Fourier transform..t' introducing the Fourier

'" 'U 'U 'U
transforms (p'), (p) and (q), w (q), and then do a Cini-
~
Touschek transformation with respect to q. Instead we shall con-
sider a simplified situation corresponding to quasi-free heavy
fermions that will justify the static limit. Then
~+::::~ (ID,.J<j' (2.5.5)
correspond approximately to the particles and antiparticles.

Hriting H for the large quark mass, we can consider an effective
interaction of the quarks with mesons ( and w ) that are in fact
~
qq bound states bound by gluon exchange forces. If ~ is quasifree
we can write
(2.5.6)
+ total divergence
(2.5.7)
The first term is of the order ).l 12M if).l is the meson mass and
negligible in the static limit. In the second term the operator
Y4 Y5-+
Y -+
-ocommutes
= ' h Y an d as a resu ltoes
W1t d not m1X. part1c' 1e
4
states with antiparticle states, so that putting 1;+ = I; (particle
wave function) we can make the replacement
.i t.,.. (.~ .)~

2M J cr l \7 <p ) (2.5.8)
1090 F. GURSEY
For the vector meson W is not independent of the independent

o
variables w. (i = 1,2,3) and we can write ig~y.~w = gwtY5o.~W
l.
Expressing small components in terms of the large components in the
free-Dirac approximation, and neglecting total divergence terms we
can make the replacement
(2.5.9)
In momentum space i~ gives the virtual space-like momentum k of the

-+
mesons, so that if we call 01/ and o~ the 10ngl.tudl.na1 and trans-
verse spin operators we have
, (2.5.11)
(2.5.12)
and the interaction term takes the approximate form
(2.5.13)
with <p and Uj.L forming an SU(2) w triplet. Woo

II
is a singlet.
If we interpret.;c. in x space, then
l.nt
(2.5.14)
(2.5.15)
(2.5.16)
(2.5.17)
(2.5.18)
with
(2.5.19)
-a
all and 0.L generate SU(2\ and combine with an internal SU(N) to
give SU(2N) Performing a Cini-Touschek-Melosh transformation in
-r w
the k direction for the kinetic term we see that the generators of
this SU(2N) commute with the transformed Hamiltonian including the
w
effective approximate interaction term.
Here it is important to note that the Yukawa coupling (2.5.3)

is invariant under the local symmetry group SO(4,1) acting on the
five-dimensional representation (, w~ ) of SO(4,1). But this
group does not leave~ invariant since Eq. (2.5.4) that is a con-
o
sequence of the choice of~ is not covariant under SO(4,1). Only
o
the non-local SU(2) is the symmetry of.:t:. = , +~. I f we
w 0 ~nt
include the internal symmetry group SU(3), then we can write a

Yukawa interaction term that has SU(6,6) invariance instead of
SO(4,1) in our simplified example. This is the group introduced
by Salam et al . As is well known, SU(6,6) cannot be a symmetry of
1092 F. GURSEY
the Lagrangian. But as far as we can separate positive and nega-

tive frequency part of ~ we have an approximate SU(6) symmetry for
w
the whole Lagrangian. In QeD where the quarks are quasi-free inside
the hadron, such an approximation is justifiable because of asympt-
otic freedom. In that sense we may say that the approximate vertex
symmetry mixing spin and internal degrees of freedom like isospin
is better understood within the QeD framework. The covariant
generalization involves the spin current densities (2.4.35) and
(2.4.39). The corresponding covariant spin operators n~v and n5~
that are conserved in the free quark approximation do not lead to

72)
a close algebra ~n general. However it can be shown that the
matrix elements of the transverse part of n5~ and the longitudinal
part of n~v between one particle states do lead to closure and to
SU(6) symmetry in a special frame.
w
since mesons are extended objects, a quark-meson coupling such
as (2.5.3) cannot be a local interaction. In general there will be
a form factor F(q), where q is the momentum transfer at the vertex.
Expanding the form factor in powers of q is equivalent to writing
effective interactions involving higher derivatives of the meson
fields. As an example we can write an effective Lagrangian with
pseudovector coupling for the pseudoscalar meson and a Pauli coup-
ling for the vector meson w~, so that
(2.5.20)
where f is a new coupling constant. We can also write
(2.5.21)
Taking the momentum it of the exchanged mesons to be space-like and

along the third (longitudinal) direction, again we find
(2.5.22)
again exhibiting the W-spin invariance of the vertex.
In the one boson exchange approximation between heavy quarks

we consider the case when the quarks emit two gluons that can turn
into a color singlet, C-even, qq bound state (effectively a pseudo-
scalar meson ) or the alternative case when they emit three gluons
that go into a color-singlet, C-odd, qq bound state (effectively a
vector meson w). Then the mesons and w will have effective
11
coupling constants (g for non-derivative and f for derivative coup-
lings) arising from the expansion of their form factors. We can
use these couplings also to describe the baryon baryon potential
73)
due to meson exchange Indeed the baryons host three quasifree
quarks each. Following the usual quark additivity assumption we
shall let one quark from each baryon interact at a time, while the
other (spectator) quarks remain idle. When the active quarks emit
several gluons in a colored state, that state cannot propagate due
to color confinement. The active quarks can make contact only
through emission of two or three gluon color singlet states which,
in turn can create mesons in intermediate states. Hence g and f
will also be effective coupling constants for the baryon-meson
interactions and will enter in the calculation of effective baryon
baryon potentials due to meson exchange. Thus, starting from a
scenario in QCD we are able to justify the Moeller-Rosenfeld model
in which baryons exchange vector and pseudoscalar mesons with
symmetrical coupling constants. Accordingly we consider in the
1094 F. GURSEY
following the effective quark-meson vertex with no renormalizabil-

ity restriction on the form of the Yukawa interaction.
The exchange between quarks will give for the Fourier trans-
-+
form v(k) of the potential
(2.6.1)
-+
where k lS the space-like momentum transfer (in a frame with k
o
= 0),
m lS the meson mass, M the quark mass (that becomes the baryon mass
when we add the masses of the spectator quarks) and f the pseudo-
scalar coupling constant. Its Fourier transform gives the potential
73)
due to , namely
(2.6.2)
where and X are functions of
x= mr (2.6.3)
5 12 lS the tensor operator
3 C'(I)
6 r ....)((5 .+)
.. (2)r
5 i2- ;: r2.
...., (,) - (2.)
(5 cr , (2.6.4)
with
- I'~ I,
.~
r r -
I
('
2 (2.6.5)
and
. -J(
X{x} = (-3' ... ..!..x 1" .X:-

..!.... ).!S.
X (2.6.6)
Similarly, the exchange of the vector meson w gives in the same

]J
approximation of neglecting recoil
(2.6.7)
In this limit we can also put
(2.6.8)
In this approximation the total potential V is
(2.6.9)
Introducing the projection operator
(2.6.10)
with eigenvalue 1 on spin singlet states, we have
(2.6.11)
Hence the potential is a superposition of two spin independent

potentials namely ~ and ps~. The tensor terms S12 in V~ and VW
are spin dependent and highly singular at x= O. If ~ and w have
the same mass and the same coupling constant these terms cancel,
resulting in a smooth wave function for the solution of the
Schrodinger equation with V as potential. Rarita and Schwinger 47 )
observed that even in the case m~ f row the potential is smooth
1096 F. GURSEY
enough for the Schrodinger equation to be soluble provided that
In order to generalize this model to include isospin we note

that diagrams considered so far violate the Zweig rule 74 ) according
to which disconnected quark diagrams are suppressed. The and w
that are exchanged In these Zweig-forbidden diagrams are not only
color singlets but also flavor singlets, i.e. they have zero isospin
or unitary spin. Another effective interaction between quarks will
occur through the exchange of color singlet but flavor carrying
qq bound states, such that q and q are associated with different
flavor quantum numbers. For example we can have
d+1.-l - d+(u,+J..)+c1..""" d.+u, (2.6.12)
where the bracketed u and d form an intermediate bound state

(TI + or p+) that lS
. exchanged between a u quark and a d quark. Again
we can call g the effective coupling constant (u d TI+) or (u d p+).

If m is the mass of the exchanged meson and M the common mass of
u and d assumed to be degenerate, the potentials resulting from
+ +
TI and p exchange will have respectively the forms (2.6.2) and
(2.6.7), so that for the equal coupling constant and equal mass
case the total potential ,li11 have the form (2.6.11). Including
all the charged and neutral states, the total u-d potential results
from the exchange of 0 mesons n, TI~ and 1 mesons w~ and
l
p~ ,
vJhere i = 1,2,3 are isospin labels. Denoting the degenerate (u d)
pair by ~(I = 1/2) we have the effective interaction term
(2.6.13)
Including the spin states, the quark doublet (particles) form

the 4-dimensional representation of SU(4) while the bound state n
-+ -+
is an SU(4) singlet. 'IT, wand p form the adjoint l5dimensional
representation of SU(4). With respect to the SU(4) symmetry of
w
the vertex, w" is the SU(4)w singlet, while the SU(4)w(15) consists
of
.
l
~Il (2.6.14)
The exchange of these mesons leads to the potential of the form
(2.6.15)
that commutes with the SU(4) generators found from ~(~(ll +~~)) and
~(~(~ +~(~). This potential is therefore spin and isospin independ-
ent in the limit of degenerate masses and equal coupling.
. ~ i
If the mass degeneracy is lifted by al 1 ow~ng 'IT and p to have
different masses, a tensor force S12 occurs in the potential between
u and d, hence, between the proton and the neutron, so that the
deuteron acquires an electric quadrupole moment. The right sign
an d .
Qagn~tu d e are found 49 ) ~f t h e P ~s
~ . tree
h to f our t'~mes h eav~er
.
than the pion, in accordance with experiment.
The SU(4) symmetry we have found is broken when we add to the

potential recoil terms associated with the one boson exchange. These
-+ -+
are proportional to L.S (spin-orbit) and Q12(quadrupole) given by
-
L-
.-+ -+
r x. I:>
-
5 .- (2.6.16)
1098 F. GORSEY
(2.6.17)
Using ~ and X defined by (2.6.6) we can write the additional

terms associated with w exchange as
(2.6.18)
++
The exchange of the pseudoscalar ~ or Q12
contributes no L.S
terms that may cancel the highly singular additional terms due to
w exchange. A Q12 term that can cancel the Q12 term can only come
from the exchange of a scalar meson a. Accordingly we can enlarge
our scheme by adding a scalar field a with the Yukawa interaction
(2.6.19)
to the interaction term (2.5.4). Then we have an additional poten-

tial due to a exchange that reads 73 );
(2.6.20)
In the total potential
V= v' + VW + yO'
the Q12 terms cancel, leaving
(2.6.21)
The term with pS (where pS is given by (2.6.l0~ is spin indep-

endent while the term in t.~ breaks this symmetry.
The M-R model without isospin is based on w and ~ exchange.

We have seen that, neglecting recoil, the model gives a spin-indep-
endent potential between the fermions. Including recoil we get a
highly singular potential with spin dependent, spin orbit and quad-
rupole terms. The extended M-R model, with the inclusion of a
scalar a-meson gives a spin independent potential broken only by
relativistic spin-orbit terms. But now, for vanishing fermion mass,
the Lagrangian exhibits a higher symmetry, namely invariance under
U(1)xSU(2) where U(l) which rotates ~ and a is a chiral group.
w
It follows that the extended model has both SU(2) and chiral invar-
w i
iance. If we introduce isovector fields TI~(pseudoscalar), p~
(vector), oi(scalar), the potential becomes
W =- (" I + "ti')c:t)) V , (2.7.1)
where V is given by (2.6.21). The new potential is both spin and

+ +
isospin independent except for spin dependent L.S terms. The
-
generators of SU(4) are
w
S = ~ (~(I) ... (;(1)), T t:= 1.2 (T~I) + -c~~)

l ')
5T i. (2.7.2)
1100 F. GURSEY
SU(4) becomes SU(6) if the isospin is replaced by unitary

w w
spin. But the Lagrangian also has an additional chiral symmetry in
the zero quark mass limit. If the masses of the u and d quarks are
neglected we have an SU(2)LxSU(2)R invariance in chromodynamics.
For free quarks, combination with spin leads to SU(4)LxSU(4)R
lnvarlance. qq states belonging to the adjoint representation will

have both positive and negative parities. Adding singlets 0 and
we consider the 32 states
En , 5', </>, w., (n =. I) 2.,3),

,
AL SL Ti. ,l
~l
n
(L = i) 2) 3) (2.7.3)
n >
where i is the isospin index and n labels the independent spatial

components of mesons with spin 1. Here , n., w pi have negative
l n n
parity while the scalar mesons (0, 0i) and the pseudovector mesons
(E , Ai) have positive parity. The scalar mesons 0 and O. are
n n l
necessary to cancel the Q12 terms due to wand p exchange respect-
ively. Thus the smoothing out of the wave functions for bound
states of light quarks (for which the recoil corrections are import-
ant) enlarges the SU(4) vertex symmetry that combines spin and
w
isospin to include also the chiral symmetry SU(2)LxSU(2)R.
As indicated before, using the additivity principle for the

quarks inside the baryons and letting one quark in each baryon
exchange a meson at a time, both the SU(4) symmetry and the chiral
w
symmetry are reflected in the potential betvleen two baryons.
Note that the cancellation the Q12 terms requires the degen-
~f
eracy of 0 with wand oi with p. This corresponds to an 0(4)

n n
subgroup of SU(4)x(SU(4) which seems to be broken less than SU(4).
SYMMETRIES OF QUARKS AND LEPTONS 11 01
He have seen that an approximate combination of spin and

internal degrees of freedom arises in effective Lagrangians where
the couplings of the qq bound states to quarks (q) or baryons (qqq)
are almost equal and the masses are almost degenerate. The
SU(4), SU(6) or SU(8) generators that close are not general cov-~n
a
ariant. They are related to the covariant operators an' anA by a
unitary transformation that is momentum dependent, hence non-local.
This unitary transformation can be worked out for the free quark
case. It serves as a model for the interacting case and works very
62-64)
well in phenomenological applications . The success of this
procedure is now partially understood by the asymptotic freedom of
quarks interacting through gluon exchange. At short spatial separat-
ion quarks are quasi free, so that their positive and negative freq-
uency parts can be separated and they can be subjected to Foldy-
Houthuysen-Tani or Cini-Touschek-Melosh type transformations. The
vertex then exhibits an SU(6) type symmetry for a quark emitting a
w
meson with space-like momentum. If the meson momentum is time-like,
the meson can only be coupled to a quark-antiquark pair, each with a
time-like momentum (in general not on the mass shell). In this case
the effective interaction Lagrangians of the form (2.5.3)and(2.5.20)
have an SU(2) (no flavor), SU(4) (isospin flavor) or SU(6) (unitary
spin flavor) symmetry with the correspondences (in the SU(4) case)
l -L
T ....... Il ) J
(2.8.1)
which give the usual SU(4) classification for the bound states qq.
The spin dependence of the one-gluon exchange between q and q
. -+(1) -+(2)
~ntroduces a a .a term that splits TI from wand p and leads to
the mass formula for SU(6)
1102 F. GURSEY
that was proposed 3 ) in 1964. The M part comes from the exact
o
5U(6) limit valid for the spin independent confining potentia1 7S ).
This spin independence is only seen in the lattice approximation in
QCD. We have found a partial justification in the Moeller-Rosenfeld
model with quarks exchanging almost degenerate qq bound states.
The 5(5+1) part comes from short range one-gluon exchange forces
as shown by de Rujula, Georgi and Glashow 3S ) who find the value of
M3 to be proportional to as' the QCD fine structure constant and
inversely proportional to the masses of the constituent quarks.
Finally the Ml and M2 terms come from quark mass difference.
The spin independent confining force can be largely simulated

. . . 76)
by an effect~ve scalar f~eld as shown by 5chn~tzer
A phenomenological bag model description of the hadron is

obtained by R. Friedberg and T.D. Lee 66 ) if the phenomenological
scalar field (regarded as a function of invariants constructed out
of two-gluon fields and three-gluon fields) is assumed to have
different vacuum expectation values inside and outside the hadron.
As shown by the same authors the large mass of the color and flavor
singlet can be understood by the contribution of the 2-gluon
annihilation diagram for qq.
Once we have this phenomenological picture for hadrons we can

get non-trivialdecay amplitude relations for mesons going into 2
mesons or baryons decaying into baryons (by emission of mesons or
photons) by using the unitary transformation between the vertex
spin operators and the covariant spin operators. The non-covariant
spin belongs to the 35-dimensional adjoint representation of 5U(6)
w
It is a function of the covariant spin operators of current algebra
and the momentum. The momentum dependence can be taken into account
77)
by consideration of the group 5U(6) x50(3) where the 50(3) is
w
associated with the orbital angular momentum exhibiting the momentum
dependence. The 50(3) part can be described by the quantum number
L= 0,1,2, etc. We have seen that in the infinite momentum frame

Y5 converges to the longitudinal spin al 1 or a3 (with 02 being
chosen as the longitudinal direction). The pion operator Y5f in
+
current algebra becomes a 3 ,. The conserved spin is obtained by a
unitary transformation of this operator. It will belong in general
to a superposition of (35) L = 0, L = 1, L = 2, etc. For a
given natrix element of the pion operator between two states (meson
states for pionic decay of mesons, baryonic state for pionic decays
of baryons) only a limited number of these representations (35, L)
will contribute. That will immediately give non-trivial relations
between meson-meson-pion amplitudes or between baryon-baryon-pion
amplitudes. y-decays or vector meson decays are handled in similar
fashion. The details of this analysis can be found in the reviews
78) 79)
of Meshkov for meson decays, Hey for baryon decays or in the
. 1es 62-64) exp 1
numerous original art1c
or1ng th
e un1tary trans f
ormat10n
between the SU(6) operators of the non-covariant constituent quarks
and the SU(6) operators of current algebra generated by covariant
current quarks. More model dependent and detailed descriptions of
the bound states and decay processes are obtained in a semi-classical
field theory of quarks bound by a confining (harmonic or linear)
potential that is from the outset taken to be spin independent 75 ).
The chiral structure is related to the approximate zero mass

of the u and d quarks and the negligibly small mass of the pion 8D )
In quark phenomenology the u and d quark masses are usually taken
to be one third of the proton or neutron masses. On the other hand,
if they are evaluated from the deviations from exact chiral symmetry
in current algebra, they are found to be of the order of a few MeV.
The two pictures can be reconciled by putting massless quarks in a
bag (resulting from an effective scalar field) of radius R, assuming
that the sphere of radius R is the boundary of regions where the
scalar field assumes different vacuum expectation values. Then,
1104 F. GURSEY
quarks behave as if they had a mass proportional to R- l which can

be adjusted to one third of the proton mass. For an attempt to
understand the zero pion mass within this phenomenological bag
model based on QCD, the reader is referred to the paper by
. 66)
Fr~edberg and Lee .
We have seen that supersymmetry ~s not necessary for the

understanding of SU(6) type phenomenological groups that combine
spin and internal symmetries. We have given arguments that may
lead to the understanding of the spin independence of the confining
long-range interaction of quarks and the spin dependent effects of
the short-orange interactions in the framework of QCD, namely the
Yang-Mills theory based on the local gauge invariance of the color
c
group SU(3). In this sense we can say that even if we do not
understand SU(6) completely we have a program for understanding it
in a QCD quark field theory.
III. UNIFICATION OF QUARKS WITH LEPTONS.

EXCEPTIONAL GAUGE GROUPS
We have seen that the phenomenological symmetries of the had-

ron~c world can be, or can be hoped to be understood in terms of a
quark substructure. The quarks carry two kinds of internal quantum
number: flavor and color. The color group, (by which we mean an
exact color group) is SU(3)c. The flavor group depends on the
number of colored quarks and .their grouping in representations of
the flavor group. All quarks may fall in one representation of the
flavor group GF or in several irreducible representations of GF ,
We now know that there are at least 5 kinds of colored quarks,
namely u, d, s, c, and b, Their space-time properties are described
by the (0, 1/2) + (1/2, 0) representation of the Lorentz group so
that in a field theoretical description they are all Dirac spinors

~. Since the separation of the left and right handed parts ~L and
~R is Lorentz invariant, we may consider separately the flavor
group representations of the left handed and the right handed
quarks.
The quark fields are assumed to obey field equations and inter-
act locally with boson fields. These fundamental bosonic fields
are of two kinds. The ones associated with the exact color group
are obtained as gauge fields arising from the local validity of
SU(3)c. They are 8 massless vector gluon fields in the adjoint
representation of SU(3)c which only interact with quarks and with
each other and consequently are responsible for the strong inter-
actions of quarks. Quarks also interact with a massless photon, a
vector gauge field associated with a U(l) group. But the photon,
unlike the gluons, does not interact exclusively with the quarks.
The same is true for the massive weak bosons (yet to be discovered)
that are responsible for the weak interactions of quarks. There
+
are at leas't 3 weak bosons. Their associated vector fields W-, Z ,
11 11
together with the electromagnetic field potential A can be inter-
11
preted as the gauge fields of the Weinberg-Salam group SU(2) x U(l)
that combines weak and electromagnetic interactions. Quarks are
also assumed to interact with scalar fields that belong to some
representation of the flavor group but are singlets under the color
group. These Higgs scalars break the flavor group spontaneously,
keeping only the electromagnetic U(l) exact. This process of
spontaneous symmetry breaking through the vacuum expectation values
of the Higgs fields gives masses to some flavor gauge bosons, for
example to the three weak bosons. It is not clear whether the
Higgs fields are fundamental or are just effective fields arising
from qq bound states. If the local gauge symmetry is extended to
local supersymmetry, then the existence of fundamental Higgs fields
1106 F. GURSEY
would be required by a symmetry principle.
The quark flavor group must include the Weinberg-Salam SU(2) x

U(l). In terms of this group we already know that the left-handed
quarks u, d, c, s are in doublet representations and the right-
handed fields ~, dR' are in singlet representations of the W-S
SU(2). There is also some mixing, namely the Cabibbo mixing between
dL and sL' so that, denoting the Cabibbo angle by e we have the
doublets
(3.1.1)
and the singlets
(3.1.2)
The transformation properties of c R' sR' b L , b R are not known

1n a definitive way. In the simplest version of the Weinberg-Salam
theory it is assumed that all left-handed quark fields occur in
doublets and all right-handed quarks are in singlets. Since the
81)
b field seems to have charge -1/3, this scheme necessitates the
introduction of a sixth quark t (top quark) that together with b
(bottom quark) forms a left-handed doublet. The t quark would have
the same charge 2/3 as u and c. Thus, we are led to the extended
scheme
, ( :~) , (3.1.3)
(3.1.4)
1n which there could be new Cabibbo-like mixings of b L with dL and
sL'
Now, the gauge-bosons associated with the flavor group inter-

-
act also with leptons whose list includes e , ).I , , and their
e ).I , ,
neutrinos \l L , \l L, and \IV The last one \l L is not on the same foot-
ing as the electron and muon neutrinos and the limit on its mass is
150 MeV. Leptons share many characteristics with quarks. They are
described by local Dirac fields and under the W-S SU(2) x U(l) they
have the assignments
(~:) ,
I-
(;:) (;~) , (3.1.5)
(e~) . (3.1.6)
Again, the definite assignments for ).IR and 'R are not known. In the
simple version of the W-S theory one assigns all the right-handed
leptons to SU(2) singlets so that we can complete (3.1.6) as
(3.1.7)
If we accept provisionally the assignments of the simple W-S

scheme, the only important difference between leptons and quarks is
given by the color degree of freedom of the quarks that forces them
to have strong interactions with gluons. Another difference is
that they carry fractional charges while the leptons have integer
charges. Thus the assignments under U(l)el are different. Finally
the quark masses are quite different from the lepton masses. But
as far as their Lorentz properties, their structureless point field
character (meaning parton behavior for quarks) and their SU(2)
flavor properties there seems to be a symmetry between quarks and
leptons.
Suppose the assignments of the simple W-S scheme are not cor-
rect but (3.1.1), (3.1.2), (3.1.5) and (3.1.6) still hold. We
1108 F. GURSEY
could also add (~R) as a singlet to (3.1.6) in order not to destroy

electron-muon universality. In this case should we abandon the
quark-lepton symmetry? It is possible that quarks and leptons are
still symmetrical in a more subtle group theoretical sense by fall-
ing in the same representations of a flavor group. For instance
the spin-1/2 baryons have a symmetrical behavior because they are
in the same 5U(3) octet. But, with respect to the 5U(2) subgroup
of 5U(3) we have two doublets (p,n) and (=0, =-), but also a singlet
(A) and a triplet (~). Then, new heavy leptons and quarks may be
necessary to complete the quark-lepton multiplet.
When leptons and quarks are put in the same multiplets of a

group G, this group will have transformations that can change lep-
tons into quarks. The group G must commute with the Lorentz group
so that the fermions can always be put in a left-handed represent-
ation. In this case ~R is replaced by
(3.1.8)
which is left-handed and defined as the left-handed part of the

charge conjugate field. An example of this unification is given
31 82)
by the 5U(5) group of Georgi and Glashow' Consider the
5-dimensional fundamental representation and the lO-dimensional
antisymmetric tensor representation of 5U(5)
AI
dR 0
A"l> 1\2
Utt -lA./\ -d.' I..
I
iLl..
,,3
-Ll.jl 0
AI (\1 \.lI-
Ce)
a~ (e)
/..lit - l-
dR3 ,
"Zo AI
0 -d~ 143
~S)= ~iO)- LlR -L4.ft L.
"e
l-
d.'I.. d2.L d1I. 0 "
eR.
-u.' _v..1 _tiJ _~ 0
e~ 1-1- I. A. (3.1.9)
These representations correspond to the decomposition
Sues)~ SU{2.) X U(I) X SU(~)C (3.1.10)
e
of the group G. The quark indices refer to color. The SU(2)weak
generators are
l ') 0 )
,-, 1(')_-- ( 0I 0
J
0)~ .
o '2 T 0 0 (3. 1. lla)
The color generators are
C-A 0)
~ lil'
I (2.)=.
(~ 0)
I ' (3.1.11b)
where Aa are the 3x3 Gell-Mann matrices for SU(3), and the electric
charge operator in SU(2) x U(l) is
Q=
3
0
0
0
I
0
0
0
I
3
0
0
I
:3
a a
0 0
0 0
()
0
==
(i
0
1(') 0 )
-4 (,-T3)
0 (3.1.12)
0 0 0 0 -I
This imbedding of SU(2) x U(l) into SU(5) leads to the

Weinberg angle parameter
Sln
. :2.6 w
(3.1.13)
as shown by Georgi and Glashow.
The remaining 12 transformations of the coset space
su(s)j SU(2.) X U(J)X SU{?" (3.1.14)
are associated with new vector bosons in a gauge theory. These new
bosons acting on ~(5)
e into -dR'
will transform e L and v L Since they
carry both lepton and quark quantum numbers, they are called lepto-
quarks. If we consider the action on ~(10) we see that some bosons
1110 F. GURSEY
will change quarks into antiquarks. Thus they will behave like
diquarks or antidiquarks. Some bosons will be simultaneously
diquarks and leptoquarks. When such bosons are emitted a quark can
turn into a lepton or an antiquark. It follows that by the exchange
of a diquark-leptoquark boson, two quarks can turn into a lepton
and an anti quark.
(3.1.15)
If two quarks of the baryon (3-quark system) undergo this change we

have
so that the proton can decay into a lepton and a meson.
It follows that the unification of quarks and leptons makes

. . . 29)
the proton unstable as f~rst po~nted out by Pat~ and Salam To
account for the observed quasi-stability of the proton the unifying
group must be spontaneously broken by Higgs fields that acquire
vacuum expectation values giving the diquark-leptoquark boson a
mass of the order of the Planck mass 1017GeV This corresponds to
-33
distances of the order of 10 cm and requires a huge extrapolation
for the validity of the present theory into high-energy regions that
will remain outside experimental exploration for the foreseeable
future. Many physicists are not prepared for such an act of faith
in wild speculations. There may be other kinds of unification that
will not involve such huge masses. However, they are required by
the simplest kinds of unified gauge theories. In the following we
shall try to accommodate this seemingly bad feature of the theory.
The (5) and (10) dimensional representations of SU(5) put the

doublets (v~, e L), (~, dL) and the singlets eR, uR and dR into
two representations of G. We may repeat this process for the next
two columns associated with (V~, ~, c, s) and (v~, T, t, b). If

there is a further SU(3) symmetry with this triplet then all basic
fermions would fit into the (5,3) and (10,3) representations of a
group SU(5) x SU(3) which is a subgroup of the simple group SU(8).
Thus, unifying groups are fairly large groups with minimum rank 4
that may reach 8 or more.
The large mass of the leptoquarks will cause large renormal-

83)
ization effects. These were studied by Georgi, Quinn and Weinberg
who showed that the renormalized values of a and sin 2e can be
s w
shifted by large amounts compared to their unrenormalized values.
A leptoquark mass of the order of 1019GeV for instance will lead to
a of the order of 0.3 and sin 2 e of the order of 0.2. Surprisingly
s w
these are excellent values that are required by the phenomenological
analysis of the charmonium system (cc bound states) and high energy
neutrino scattering experiments. In other words we would have found
again a mass of the order of the Planck mass if, irrespectively of
the proton life-time we had tried to unify the fine structure con-
stants of strong and electromagnetic interaction (a and a ) or if
s e
we had conjured renormalization effects that will bring the unrenor-
malized 3/8 to the observed 0.2 for the Weinberg parameter. Hence,
it may be that the bad feature of the grand unification may be a
useful one in correlating a and a or in explaining the observed
s e
value of the Weinberg angle. This makes us take the superheavy
mass scale more seriously. Are there other bad effects of such a
84}
mass? Appelquist and Carazzone have shown that for most other
purposes the superheavy particles decouple from the rest and cause
no harm.
There remains one other question. The Planck mass is the mass
that comes in Quantum gravity. Hence, a theory involving such a
large mass cannot ignore gravitation. The large mass scale may make
sense ~n a superunified theory of strong~ electromagnetic, weak and
1112 F.GURSEY
gravitational interactions. We are far from such a theory. First

of all, pure gravitation is not an ordinary renormalizable gauge
field theory. Its unification with the electromagnetic field
requires extended supergravity based on local supersymmetry. There
are attempts to unify gravity with internal gauge groups beyond
U(l), namely SU(2), SU(2) x SU(2) and perhaps orthogonal groups up
to SO(8). But it is not at all clear how gravity could be unified
with a gauge group as large as SUeS) or larger.
Until new ideas for the unification of gravity or supergravity

with the weak, electromagnetic and strong interactions are intro-
duced we shall confine ourselves to the problems of the more modest
unification (called grand unification) of weak, electromagnetic
and strong forces.
An immediate extension of the SUeS) theory would attempt to

put the electron family of basic fermions (v~, e, u, d) into a
single irreducible representation. This is achieved by embedding
SUeS) in SO(lO) and selecting the l6-dimensional spinor represent-
ation of SO(lO). Under the SUeS) subgroup we have the decom-
position
16 ::; I .,. 5 + 10 (3.1.17)
It follows that, with the addition of a 2-component neutral

e (e) (e)
lepton Cl. L , the1/J(S) and lji 10) of (3.1.9) fit in a single spinor
representation of SO(lO). We can repeat this process by consider-
ing three different spinor representations of SO(lO) to fit all the
known basic fermions in 16+16+16.
Another possibility is that the e and f.l families are in 16-

dimensional representations of SO(lO) and the T family in a
representation with a different dimension. Alternatively, with the
addition of other fermions than t to the known ones we may have
a different distribution of fermions among different representations

of 50(10). The E6 scheme will be such a scheme.
The smallest groups in which SU(5) can be embedded are SO(lO) and
SU(6), both of rank 5. The SU(6) assignments have been recently
studied by Segre et al;4). In order to cancel Adler-Jackiw-Bell
anomalies, two different representations of SU(6) are required,
namely two six-dimensional and one IS-dimensional representations.
These would correspond to the
(~,z) + (is,I) (3.1.18)
representations of SU(6) x SU(2). A possible assignment is the

following:
o 1\ 2 .'
-u R. -dL IA,. bl-
I I
I
I
I
"1 "I .2. 1 .2

-u. R 0 tAR -el ... 1A.j...
I
b",
11.1
u.~ -u"
",
O-d?'t.. 1
1.1. .. 1
I ~
0;..
I
. .z. 1\ A
~b') == dL. d .. d~L.. 0 eR. IN
I R
.(! I
r::;.1- /
(3.1.19)
Note that, with only one 6-dimensional representation we

34)
would have anomalies leading to a nonrenormalizable theory To
accommodate the muon family we must have another copy of the three
. h b . h e)
representat~ons s own a ove, w~t a ,~
Q (e)
, ve, V , e, L, u,
L
(~) (~) ~ M
d, b) replaced by (a ,S ,v, v , ~, M, c, s, h). Thus, the
SU(6) generalization does not require a charge 2/3 top quark, but
a new charge -1/3 heavy quark h. The embedding of SUeS) is shown
in dotted lines, so that the SU(2)weak is embedded in SU(6) via
1114 F. GURSEY
SUes) as before. This allows us to read the weak doublet structure
::
from (3.1.19). The doublets are
( ), (;:'), (;~), (~:), (3.1. 20)
with all the rest being singlets. Similarly, for the muon family:
(3.1.21)
Thus, in the SU(6) unification scheme the simple W-S theory is

valid for e, V, u, d, c, s but not for the heavy leptons and heavy
quarks. The decay of the b proceeds only through neutral weak
currents and Cabibbo-like mixings of the b with sand d. Soon we
may be able to tell whether the known basic fermions come in 3
copies of the Sand 1U of SUeS) (or 3 copies of the spinor re-
presentation of SO(lO) ), or, alternatively, as 2 copies of the 6,
6 and ~ of SU(6). Both schemes require 6 quarks. In the SUeS)
approach the sixth has charge 2/3 and is in a weak doublet while
in the SU(6) approach, the sixth has charge -1/3 and is in a weak
singlet.
-
The b quark, hence the weak decay modes of bb and mesons
involving a single b, will be different in the two schemes. Further-
more the V-A character of T decay tells us that if TR is in a weak
doublet as in (3.1.20), then, the neutral companion NR of TR which
has V + A interactions must have a large mass so that at low ener-
gies we only see the V-A decays.
~Je have seen that the unification of weak, electromagnetic

and strong interactions leads to simple groups that admit
SU(2) x U(l) x SU(3)c as a subgroup. They could in principle belong
to one of the four families of Lie groups. The first three famil-
ies have Lie algebras associated with antihermitian matrices with
elements that are in one of the three associative normed composition

algebras R (real numbers), C (complex numbers), H (quaternions).
For any n there are n x n antihermitian matrices over R (in this
case the matrix is antisymmetric), C or H (in this case hermiticity
is defined with respect to quaternionconjugation that changes the
sign of the 3 imaginary quaternionic units e. = -icr. , with
J J
j = 1,2,3 and cr. denoting the three Pauli matrices). The Lie alge-
1
bras associated with R, C and H are respectively the Lie algebras
of the orthogonal, unitary and symplectic groups. Since n is
arbitrary each of the three families is infinite. The fourth
family is associated with the fourth Hurwitz algebra, namely the
normed composition algebra of octonions. Octonions have seven
imaginary units and their algebra is neither commutative nor
associative. This non-associativity results in the fourth family
to have only five members, namely the exceptional groups
G2 , F4 , E6 , E7 ,and ES' The subscript denotes the rank of the group.
These groups have many intriguing properties.
a- They are all connected with octonions.
b- They all admit SU(3) as a subgroup. We may connect this SU(3)
which is common to all five with the color SU(3) of quarks.
c- The exceptional family is not extendable so that there is only
a finite number of possibilities for the symmetries of quarks
and leptons if these groups are identified with the unification
groups. The number of fundamental degrees of freedom is then
finite in an essential way.
d- These groups are connected with new kinds of projective geo-
metries or higher unique geometries that are non-Desarguesian.
It is therefore tempting to take gauge theories based on these

groups as models for unified theories. Clearly G2 is too small: it
has color but no flavor. F4 has SU(3) x SU(3)c , hence
SU(2) x U(l) x SU(3)c as a subgroup. Thus, it could be a candidate.
1116 F.GORSEY
Its fundamental representation is 26-dimensional and real. Under

SU(3) x SU(3)c it has the decomposition
26= (8,1) + (3) 3) -+ (3, :3 ) . (3.2.1)
Its adjoint representation is 52-dimensional and decomposes as
(3.2.2)
Identifying the color singlet with leptons in the fundamental

representation, a (26) describes 8 left-handed leptons, three left-
handed colored quarks and three antiquarks. To accommodate the
known basic fermions we need two complex or 4 real (26) represent-
ations. The bosons consist of 8 color gluons, 8 color singlet
bosons (1 photon + 4 weak bosons + 3 heavier weak bosons or super-
weak bosons) and 6 colored leptoquarks (which are simultaneously
diquarks) and their conjugates.
The next group is E6 which has the following maximal sub-

30 85)
groups'
a) E6 ) SU(3) x SU(3) x SU(3)c

b) E6 ) 50(10) x SO(2)
c) E6 ) SU(6) x SU(2)
d) E6 ) SU(5) x U(l) x SU(2),
I t follows that all the unifying groups reviewed in the pre-

ceding section, namely SU(5), SO(lO) and SU(6) occur as subgroups
of E6' Its fundamental representation is 27-dimensional and is
complex. Thus all the basic fermions can be accommodated in two
(27)-plets, one corresponding to the electron family and the other
to the muon family, The adjoint representation is 78-dimensiona1
with the SU(3) x SU(3) x SU(3)c decomposition
78::: (8, I, IC )+(/,8, ,C.)+ (1,1,'0')+ (3)3,3c.)+(~,35C)'(3.2.3)

The 27-plet decomposes as
2.1 == ("3, 3) Ie) + ( 3) I ;3 c; ) of (1,:3)"3') , (3.2.4)
with the conjugate decomposition for the 77. Thus the 27 describes
9 leptons, 3 colored quarks and 3 colored antiquarks. Two (27)
describe 18 two-component leptons with the charged sector corres-
ponding to e, T, ~ and M, and 6 colored quarks with charges corres-
ponding to u, d, b, c, sand h. To see how the W-S group is embed-
ded in E6 we consider the decompositions b, c and d
b: [SO(lO)J: 27 = 1+10 + 16, (3.2.5)

c: [SU(6)xSU(2)]: 27 (6,2) + (15,1), (3.2.6)
d: [SU(5)xSU(2)xU(1)] ~ 27 = (1,2)+(5,2)+(10,1)+(5,1). (3.2.7)
Thus, we find automatically two (6) and one (15) of SU(6) and
(TO) and (5) representations of SU(5) with additional (5) and (1)
representations. The SO(lO) decomposition includes the spinor re-
presentation 16, but there are also additional (10) and singlet
representations. It follows that the SU(6) assignment (3.1.19) ~s
also valid for the (27) of E6 The SU(5) assignments can also be
read off after dividing the representations by the dotted lines as
shown. The weak SU(2) doublet structure is therefore given by
(3.1.20) and, for the muon family by (3.1.21). This assignment is
. 1e
compat~b . h
w~t .
exper~ments so f ar.87) Th e d'~scovery 0 f '
par~ty
violation in atomic physics is necessary for this assignment. Note

that if the roles of e and T are switched, then we have conservation
of parity ~n atomic physics with NR being now the heavy neutral
companion of e R
In terms of the SU(3) x SU(3) flavor group the leptons fall in

a (3x3) matrix that represents the (3,3). We have
(3.2.8)
1118 F.GURSEY
The first two columns are the weak leptons doublets. The (3,1)
quarks and the (1,3) antiquarks are
(3.2.9)
with the first two lines of the quark triplet being in weak doublets.
Thus the weak SU(2) is in the first flavor SU(3) group.
Similarly for the muon family we have the SU(3) x SU(3) decom-
positions
(3.2.10)
(3.2.11)
For details the reader is referred to the relevant papers: 3 ,86)

The group E7 is pseudoreal like SU(2). It has a 56-dimensional
complex fundamental representation which is equivalent to its com-
plex conjugate. Under E6 it has the decomposition
56 = (1) + (1) + (27) + (27). (3.2.12)

Hence, the muon matrix must go in a 27. The adjoint representation
is 133 dimensional. It also has the maximal subgroup
E7 ) SU(6) x SU(3)c
and the leptons fall into the 20-dimensional representation of
. . d . 88)
SU(6). The E7 model has been stud1ed 1n eta1l. It leads to
eR and PR being in weak doublets. uR and cR are also members of
weak doublets. Such an assignment does not seem to be compatible
with new neutrino experiments. Also the unrenormalized Weinberg

parameter is 3/4, which is too large.
The group E8 has no small dimensional representation. We have

to assign the basic fermions to the 248-dimensional adjoint re-
presentation which is too large for the presently explored range of
high energy physics. If the number of quarks grows beyond 6 this
ultimate exceptional group might be worth studying. Another inter-
esting feature of E8 is that it can be used in a supersymmetric
scheme since the 248 bosons can be transformed into the 248 fermions
by supersymmetry.
To summarize this section the group E6 appears to be the best

candidate with the assignment (3.1.9). It has a new lepton-quark
symmetry different from the usual one-to-one correspondence between
lepton and quark doublets. Since it leads to ~ew predictions we
may soon learn more about its validity.
The symmetry breaking must be brought about by several Higgs

fields. The first Higgs X must separate the leptoquarks and perhaps
also some color singlet gauge bosons from a family of bosons that
include the photon and the weak bosons. Such a Higgs field develops
a superheavy vacuum expectation value M of the order of the Planck
mass. Now, if X had Yukawa coupling to the fundamental fermions
it would make some of them superheavy. In a reasonable scheme X
must not have Yukawa coupling with the representation F of fermions,
hence must not be contained in the direct product F N F. We need
other Higgs fields ~, ~' which have Yukawa interactions with F,
which are therefore contained in F N F and give masses to leptons
and quarks as well as to the bosons which are not superheavy.
1120 F, GURSEY
In the E6 scheme we have

27 x 27 =1 + 78 + 650, (3.3.1)
(27 x 27) 27 + 351, (3.3.2)
sym.
(27 x 27) , 351' (3.3.3)
ant1sym.
Hence, a good candidate for X with the superheavy vacuum
expectation values is the adjoint representation. Since the'Yukawa
coupling of a scalar field is only possible to the symmetric com-
bination of (27 x 27), the representation 351' has no Yukawa coup-
ling either. It could be another candidate for X.
The only candidates for ~ with the vacuum expectation values
<100 GeV are the 27 and 351 representations.
Let us write the color singlet parts of these representations.
The color singlet part of X78 in the SU(3) x SU(3) x SU(3)c
decomposition contains
(3.3.4)
Its 50(10) decomposition contains

SO(lO) : 78 = 45 + 16 + 16 + 1.
Its SU(5) deco~position contains the adjoint (24) of SU(5).

, 3 1)
Georg1 and Glashow and Buras, Ellis, Gaillard and
Nanopoulos 82 ) use the (24) of SU(5) for superstrong breaking, which
is in line with our choice of (78) for E6
For ~35l and ~27 we have the following decompositions

SU(3) x SU(3) x SU(3)c:
(351) , = (6, 6, lC) + (3,3, lC). (3.3.5)
col. s1ngl.
(27) ,= (3, 3, lC) (3.3.6)
col. s1ngl.
The SO(10) decomposition of 351 includes the (10) of SO(lO)
and 5 of SU(5). In the SU(5) scheme the authors cited above use
(5) for the smaller mass breaking.
Finally we consider the SU(6) subgroup of E6 Under

SU(6) x SU(2)
(35,1) + (1,3) + (20,2),
(105,1) + (84,2) + (21,3) + (15,1),
(15,1) + (6,2).
The SU(2) singlets are (35) and (15) of SU(6). These are just
the Higgs multiplets considered by Segre and Weldon.
We conclude that the minimal Higgs fields necessary for the

E6 model are X78 and ~35l' The ~27 could also be necessary for
the fine structure of masses. Hence the E6 scheme contains all
the other successful grand unification models based on
SU(3) x SU(3) x SU(3)c , SU(5), SO(lO) and SU(6).
There are some problems associated with this kind of symmetry

breaking scheme. These problems are shared by all grand unification
theory. vIe simply list a few:
a- Lack of naturalness in the hierarchy as defined by Weinberg
and Glashow~9) This problem is discussed by Buras et al~2).
b- It is possible to exhibit a set of vacuum expectation values
in the 78 and 351 (or in (1,8) + (8,1) + (6,6) + (3,3) for
the SU(3) x SU(3) flavor subgroup) that gives a reasonable
86)
mass breaking. It is more difficult to show that such a
set of vacuum expectation values provide an absolute minimum
of the Higgs potential for some range of parameters.
c- Both one-loop and two-loop renormalization effects are
2 90)
important for getting accurate values of a
s
and sin e.
w
These calculations are in progress.
d- One has to show why the u and d masses are so small and how
many neutrinos are strictly and naturally massless.
e- One should relate the Cabibbo mixing to other parameters of
the theory and show why leptonic mixings are so small.
1122 F. GURSEY
IV. OCTONIONIC QUANTUM MECHANICS AS A POSSIBLE BASIS FOR

INTERNAL SYMMETRIES
In the first part of these lectures the symmetries of the

directly observed hadronic world and strong interactions led us to
the colored quark substructure. The second part showed how the
properties of weak and electromagnetic interactions help us to
unveil a deep symmetry between the directly observed leptons and
the partly observable quarks. These fermions seem to be represented
by local fields that belong to multiplets of a single unifying group
G that is valid locally. What distinguishes this group G? What
is so basic about the exact color group? In trying to answer these
questions we move to the next sphere of abstraction and see if the
unification of strong, electromagnetic and weak interactions points
to the geometrical properties of a deeper Hilbert space structure.
During the ebullient first decade after the birth of Quantum

Mechanics, the grand masters of the period explored all the pos-
sibilities for the extension of Quantum Mechanical concepts and
91)
methods. In 1933 Jordan suggested a new formulation of Quantum
Mechanics based on what became to be known as Jordan algebras. In
the standard formulation,states are represented by vectors (kets)
in a Hilbert space, while observables are represented by Hermitian
matrices acting on the states. In the Jordan formulation an alge-
bra of observables is defined so that observables can be combined
to give other observables. For instance, in classical mechanics
if p and q are observables, so are p2, q2 and pq. If P and q are
replaced by hermitian matrices P and Q, then only the combination
P.Q = Q.P = 1 (P Q + Q P) (4.1.1)
is a hermitian matrix, hence an observable representing the classic-
al pq. The symmetrized product (4.1.1) is the Jordan product of
P and Q. Unlike ordinary matrix multiplication which is non-
commutative but associative, the Jordan product is commutative but

not associative. The associator
[A,B,C] = (A.B).C - A.(B.C) (4.1.2)
is antisymmetrical in A and C and satisfies the Jordan identity

LA, B:> A 2 ] ;=. 0 , (4.1.3)
which makes the algebra of observables power associative so that

pn is an unambiguous expression that represents the classical
n
observable p .
In the Jordan scheme the states are also represented by
certain hermitian matrices P
ex
associated with the kets la>.
They are the projection operators
(4.1.4)
that project the state la>. They satisfy the relations

2-
Po< == ~ Tr~ == 1 (4.1.5)
Now, all the measurable quantities in Quantum Mechanics are

transition probabilities of the form
(4.1. 6)
where ~ is a hermitian operator and

(4.1.7)
~s the matrix element of the observable ~ between the states la>

and Is>. We can wri te 92)
(4.1.8)
We have to show that the expression between brackets can be

expressed purely by means of the Jordan product. Let
U(Q)~ :=Q~Q. (4.1. 9)
We have
TT~~ (Q) - Tr {~. U(Q)ff} ~ (4.1.10)

1124 F. GURSEY
where the dot refers to the Jordan product.
Furthermore we have the identity
{ ABC} :: i (ABC + CBA) ~ (A.B).C+ A.{B.C)-(A.C).B, (4.1.11)
so that as a special case
(4.1.12)
It follows that
U( Q) p~ =- {Q~Q} (4.1.13)
where the right-hand side is expressed by the Jordan product only.

Hence the absolute square of the matrix element Q is
as
no(~(.Q)= Tr (~. U (.Q)~)= Tr(~. fQ~Q})=T.. (U(Q)e.~). (4.1.14)
. 93)
In 19 34, Jordan, von Neumann and W1gner showed that the
Jordan algebras are always in one-to-one correspondence with the
ordinary matrix algebra over complex numbers with one exception.
8
There is an exceptional Jordan algebra J 3 of 3 x 3 matrices over
octonions which are hermitian with respect to octonion conjugation.
They have the form
0(
J
(
= ~ (4.1.15)
where the bar denotes octonion conjugation and a, S, yare se1f-

94) 8
conjugate. Then Albert proved that J 3 was the only exceptional
Jordan algebra.
The octonions a, b, c have the form
(0(=1,' .. ~7), (4.1.16)
where the e are octonionic imaginary units with the multiplication

a
law
(4.1.17)
The coefficients S are antisymmetric in (a,S,y) and non-

" a y " 95)
zero and equal to un~ty for the seven comb~nat~ons
(123), (246), (435), (367), (651), (572), (714) , (4.1.18)
obtained by cyclic permutation of the triangle (1 2 3) on a circle

with seven points labeled in the order (1 2 4 3 6 5 7).
The conjugate octonion is
>
(4.1.19)
and the norm

7
N (lL)= eta = a~ +2: (a(){/"
01:: I
(4.1.20)
has the property
N(Q.b)= N(a.)N(b). (4.1.21)
The octonion algebra is neither commutative nor associative,

but because of (4.1.21) is a normed composition algebra which
together with real numbers, complex numbers and quaternions form
the only normed composition algebras (the four Hurwitz division
algebras). The algebra is alternative, as the associator
[a..bcJ:: (~b)c -a.(bc) (4.1.22)
is antisymmetric in a, band c. The scalar part of the octonion a

is defined as
(4.1.23)
The diagonal elements a, S, y of J in (4.1.15) are scalar in

this sense.
A complex octonion is still of the form (4.1.19) but with

complex components. The resulting complex J is still an element of
an exceptional Jordan algebra although the complex octonions do not
form a division algebra.
To construct the equivalent of projection operators we

1126 F. GURSEY
introduce the Freudenthal product 96 )
J, X J 2 := J2. X J r = J, . Jz. - ~ J, Tr J2. - Jz. Tr J 1

-i I (TrJ,.J 2 - TrJ. Trjz.). (4.1.24)
The expression
Det (J) = -j Tr (J . (J X J)) (4.1.25)
1S the determinant of J.
Matrices PEJ 8 which obey the conditions
3
? X p:= 0 (4.1.26)
and
Tr P =1 (4.1. 27)
can be shown to satisfy also
) (4.1. 28)
so that they are projection operators. They have the form
( 0(1) (i);, ()(2. ex 3)

:p =01.
0(2
(4.1.29)
O(a
where a l , a 2 , a 3 are three octonions one of which is self-conjugate.
(4.1.27) is satisfied if
(4.1. 30)
P defined by (4.1.29) is a generalization of the projection

a
operator (4.1.14) in the associative case. Hence matrix elements
of an operator J can be defined analogously to (4.1.14) by
(4.1.31)
and we have a new quantum mechanics in a 3-dimensional octonionic

Hilbert space. This is the exceptional Quantum Mechanical space
discovered by Jordan, von Neumann and Wigner. In this case the
octonions are real, so that the probabilities ITaS are positive
definite. If we take the special case J = 1, then
(4.1.32)
corresponds to the square of the transition amplitude <a[S>.

97)
If we put
(4.1. 33)
then, d has a very simple geometric interpretation. The state

as
P may be associated with a point with homogeneous coordinates
a
aI' a 2 , a 3 with the constraint (4.1.30). The inhomogeneous
coordinates
-.
X,:=, 0<,CX 3 , , (4.1. 34)
when a 3 is self-conjugate, label a point in the two-dimensional

. . proJect1ve
octon1on1c .. 1
pane as s h own by Mou f ang 98).1n 1933. Then 1t
.
is easy to show that d defined by (4.1.33) is the non-euclidean

as
distance between the points P a and P s in the projective octonionic
plane.
Hence, in the octonionic Quantum Mechanics, a point represents

a state associated with the projection operator P and the distance
a
between two points is related to the transition probability between
the corresponding states.
The line in the projective plane defined by the points P a and

P s is associated with
(4.1.35)
that obeys
(4.1. 36)
Lines and points obey all the postulates of projective geometry

but do not satisfy the Desargues theorem. Hence the new Quantum
1128 F. GURSEY
Mechanical space corresponds to a new non-Desarguesian projective

geometry.
The condition for a state jy> to be the superposition of states

la> and 18> in quantum mechanics can be reexpressed in geometrical
language and in terms of the projection operators P ,P and P .
Y a 8
Geometrically, the point y with homogeneous coordinates Iy> lS on
the line determined by the points a and 8. It follows that the
determinant of the matrix with columns la>, 18> and Iy> vanishes.
It is easily shown that this condition is equivalent to the relation
Tr [B .(p~ X ~)} = 0 (4.1.37)

92 )
which uses only projection operators and the Jordan product. In
this form the geometrical property can also be generalized to octon-
ionic Quantum Mechanics, so that (4.1.37) expresses the condition
for the state y to be a linear superposition of the states a and 8.
The Desargues property can now be rephrased in Quantum Mech-

anical language as follows.
Geometrically we have a point Sand 3 points ABC not on a

line. Take a point A' on the line SA, a point B' on the line SB
and a point C' on the line SC such that A', B', C' do not lie on a
line. Now the lines AB and A 'B' intersect at a point C". Similar-
ly BC and B'C' intersect at A" and CA and CIA' intersect at B".
The geometry is Desarguesian if A", B" and C" lie on the same line.
The Desargues property holds in a projective geometry if the pro-
jective plane has dimension d>2. In 2 dimensions it holds if the
coordinates of the point are real, complex or quaternionic, i.e.
if they belong to an associative algebra. On the other hand in the
Moufang plane, the coordinates of a point are octonionic. Because
of the non-associativity of octonions, there exist non-Desarguesian
configurations.
In Quantum l1echanics we take states corresponding to the points

S, A, B, C. We superpose S and A to obtain A' and proceed similarly

for B' and C'. We now construct a state C" which is simultaneously
a superposition of states A, B and also of states A', B'. We con-
struct states A" and B" in the same way. If the states A", B", C"
are linearly related we have the Desargues property. If not, we
say that we have a non-Desarguesian Quantum Mechanics. It follows
that the octonionic Quantum Mechanics of Jordan, Wigner and
von Heumann is non-Desarguesian. It represents a completely new
kind of Quantum Mechanics in a finite dimensional Hilbert space in
which the superposition principle is modified for states correspond-
ing to non-associative triples.
Let us mention how two other features in Quantum Mechanics can

be rephrased in the Jordan formulation and hence generalized to
octonionic Quantum Mechanics.
92)
The first concerns the insertion of a complete set of states.
If I denotes the unit matrix, we have
I=- 2:
L
IL><~I ==:. ~
L
P:
L
, (4.1.38)
so that the relation
< ex I ~ > - '2 <.0< l I f. (4.1.39)

can be transcribed as
(4.1. 40)
which is also valid in octonionic quantum mechanics.
The second is related to the Jordan formulation of compatible

observab1es. In the associative case we have an identity that
relates the associator to the double commutator, namely
[AJ BJ =-4[(A,BJ,J] J
(4.1.41)
so, that if J is arbitrary and A, B, C are' hermitian the equation
[A J J3 J = 0 (4.1.42)
1130 F. GURSEY
implies
[A, BJ ;;: O. (4.1.43)
Since (4.1.41) does not hold in the octonionic case, we define

the compatibility of the observables A and B by eq. (4.1.42) for
arbitrary J. This is a condition which only involves the Jordan
product of octonionic observables.
2. ~~~~E!~~~~_~E~E~E!i~~_~!_!~~_~~~~E!i~~~1_9~~~!~~_~~~~~~i~~-~!42~
~~~~!~1i~~!i~~_E~_9~~E1~~_~~!~~~_~1g~~!~~_~~62~
In the case of the usual Quantum Mechanics transition amplitudes

are invariant under unita~y transformations as
(4.2.1)
when
t
(IX'>= U/oc> , 1/3 1
) = UI~>, UU =t (4.2.2)
Then the projection operators transform as

, t , t
Poc: :: UP()I. U , Pfl ::: UP/3 U . (4.2.3)
The observables Q that are linear combinations of projection opera-

tors also transform in the same way
Q1::UQu t , (4.2.4)
so that the Jordan product Q of two observables Ql and Q2 also

transforms like a projection operator, since
n 1:= Q; . S2~ = ~ (Q; .Q~ + .Q~ S2/)
I t
=z:U(Q, Sl 2.+ S2 2S2I)U :: US2U T (4.2.5)
It follows that, in a n-dimensional Hilbert space, with n x n

hermitian matrices associated with observables and projection oper-
ators for states, the automorphism group of the Jordan algebra of
observables is U(n) or SU(n).
In order to find the invariance group of octonionic Quantum

Mechanics we must find the automorphism group of the exceptional
Jordan algebra. This was shown to be the group F4 by Chevalley and
99)
Schafer more than a decade after the discovery of exceptional
Jordan algebras.
The infinitesimal action of F4 on an element J of the Jordan

algebra can be written simply.
If HI and HZ are traceless octonionic hermitian matrices we

have 100 )
(4.2.6)
The transformation property of the projection operators Pa for

states a is obtained by putting J = Pa. Let us show that this
gives the unitary group in the associative case. Let
(4.2.7)
Then, using (4.1.41) we can write
bJ= L[H,JJ (4.2.8)
Exponentiation gives
J': J + [i.H, J] +r!L~H,(iH,JJJ+ ,.,
= Cexp~H)J ex.p(-iH), (4.Z.9)
which shows that J is transformed by a unitary matrix. In the
octonionic case the finite transformation of F4 is given by the
series
(4.2.10)
which only involves the Jordan product and can no longer be written
in the form (4.2.9). When HI involves only one octonion and H2 is
a purely scalar matrix then (4.2.10) can be integrated in the form
(4.Z.9) with iH replaced by an antihermitian octonionic matrix
involving one octonion only. It is seen that the full group is
1132 F. GURSEY
determined by the traceless hermitian matrices Hl and H2 so that

it has 52 parameters. The invariants under the F4 transformation
are
r1 -::: Tr J ) (4.2.11)
12 :::. Tr J2 , (4.2.12)
I 3 :;: De t J :::. ~ Tr (J . J X J) . (4.2.13)
An irreducible representation of F4 is obtained by taking

o. It corresponds to traceless Jordan matrices.
We have seen that octonionic Quantum Mechanics based on real

octonions provides us automatically with a finite Hilbert space
with F4 symmetry. Since F4 has SU(3) x SU(3)c as a maximal sub-
group we have a fundamental justification for the color degree of
freedom and the SU(3) flavor:0 1 ) Under this group we have the
decomposition (3.2.1). The color singlet part which is a SU(3)
flavor octet lies in an ordinary quantum mechanical space with
SU(3) symmetry involving one of the octonionic imaginary units
while the colored degrees of freedom involve the remaining six
imaginary units.
The behavior of various states under the color group is best

seen if we use split octonion units defined by95)
Lto ;: I (I + i. e7) , 1.(!=-1. ( I - i. e,),

2
(4.2.14)
u..j:::' ~ (ej+ ieh3)' ~ =.!. (e-i.e ), (4.2.15)

j 2. J H3
(j.~ /) 2,3)
The automorphism group of the octonion algebra is the 14
parameter exceptional group G2 The imaginary octonion units
e a (a = 1 . 7) fall into its 7-dimensional representation.
Under the SU(3)c subgroup of G2 that leaves e 7 invariant u o

and u* are singlets while u. and u~ correspond respectively to the
o _ J J
representations (3) and (3).
The mUltiplication table can now be written in a manifestly

SU(3)c invariant manner:
" iI
..... 0 ...0
It = 0, (4.2.16)
(4.2.17)
(4.2.18)
(4.2.19)
together with the complex conjugate equations. Here one sees the
virtue of octonion multiplication. If we consider the direct
products
'3 x 3:::.3+6, (4.2.20)
'3X3 t+~ (4.2.21)
c
for SU(3) , then these equations show that octonion multiplication
gets rid of (6) in 3 x 3, while it gets rid of (8) in 3 x 3.
Combining (4.2.18) and (4.2.19) we find
(tti Uj) Uk :::. - t:i.jk 4: (4.2.22)
Thus, the octonion product leaves only the color singlet part
in 3 x 3 and 3 x 3 x 3, so that it is a natural algebra for colored
quarks.
We can now consider the general element F of the Jordan alge-

bra with complex components. It can be decomposed as follows
F = uoL + tA.. o* LT + Uj* Q j;- Uj R*j ' (4.2.23)
Here L, Q., R. are 3 x 3 complex matrices, T denotes transposition

J J
and Q. and R. are antisymmetric so that
~ ~
(4.2.24)
If we associate L with the (3,3) representation of a group

SU(3) x SU(3), Q. with (3,1) and R~ with (1,3), then, together with
J J
the color index j we find that F has the SU(3) x SU(3) x SU(3)c
1134 F. GURSEY
decomposition
(4.2.25)
Comparison with (3.2.4) tells us that F is the 27-dimensional

representation of the exceptional group E6 The color singlet part
L can be associated with the lepton matrix L of (3.2.8), while Q.
J
and R~ respectively are associated with left-handed quarks and
J
right-handed antiquarks (3.2.9). It follows that lepton and colored
quark fields can be combined in a complex Jordan matrix of the form
(4.2.23) which is hermitian with respect to octonionic conjugation
only, so that
-F T= U*LT+
o U.
0
L- u~QJ:r-u.f(~T=
J J J
F (4.2.26)
The i7 representation of E6 corresponds to F*.
The E6 transformation of F involves three traceless real oct-

30)
onionic Jordan matrices HI' H2 , H3 and we have
(4.2.27)
This shows that F4 is a subgroup of E6 and that E6 has 3 x 26

= 78 real parameters. The Freudenthal product of Fl and F2 projects
out the 27 representation out of the symmetric product of 27 x 27,
so that we can write
(4.2.28)
Another E6 invariant operation is the triple product defined

by (4.1.11), so that
(4.2.29)
are also invariant if F l , F2 , F 3 , F4 transform like (27). Finally

we can construct the invariant
(FJ' F2.) (4.2.30)
It follows that, with three Jordan matrices F l , F 2 , F3 we can

associate the invariant
(4.2.31)
Given one F, TrF is not E6 invariant. But we can construct

4 invariant real quantities 1 2 , 1 3 , I~ and 14 defined by
I2.=(F, F"*), 13 +LI'3==(F, FJ<F).=3])e.tF, (4.2.32)
14 = (F )( F, F* X F'It'). (4.2.33)
A geometry which generalizes the projective geometry of the
Moufang plane can be based on the complex matrices F. It is called
.. 1 102) .
t h e geometry 0 f comp 1ex octon10n1C panes. A genera11zed p01nt
(or state) is defined by S such that
SxS o (4.2.34)
The distance d 12 between points Sl and S2 or the transition

probability IT12 is given by
lTi2. =' eo/d.,2. = (5" 5 2 )= Tr (5( . S},-) (4.2.35)
and is E6 invariant.
It is possible to associate idempotent projection operators

with such states and generalize the quantum mechanical formalism.
The geometry is more complicated than the Moufang geometry and all
its quantum mechanical implications have not yet been worked out.
However, the existence of this E6 invariant exotic geometry and its
close correspondence with the phenomenological symmetries of quarks
and leptons as reviewed in the preceding lecture provides a strong
motivation for the reformulation of the properties of the complex
octonionic planes in purely quantum mechanical terms.
If it turns out that F4 or E6 describe correctly the internal

symmetries of fundamental fields we may seek the origin of these
symmetries in the properties of unique finite Hilbert spaces
.
assoc1ate dw1t
h geometr1es.
exot1c . 86)
1136 F. GURSEY
Let us briefly review the landmarks that stand out in this

exploratory journey through the intricate symmetries of elementary
particles.
The symmetries of the hadronic spectrum and the hadronic decays

have uncovered a colored quark substructure. Weak and electromag-
netic interactions showed us that quarks behave like leptons and a
local field theory of both leptons and quarks makes sense. Strong
interactions are well described by a local gauge theory based on
the exact color group while weak and electromagnetic interactions
are unified within a gauge field theory of spontaneously broken
local flavor group. The symmetries between leptons and quarks led
us to the notion that these fundamental fermions belong to the same
multiplet of a unifying group. The successful candidates for such
a unifying group have turned out to be subgroups of the exceptional
group E6 On the other hand, the only non-trivial generalizations
of the Hilbert space of Quantum Mechanics and the algebra of obser-
vables involve algebraic and geometrical structures connected with
the exceptional groups F4 and E6 These unique and intrinsically
finite structures exhibit an exact color symmetry originating in
octonions that go in the building of these exotic structures.
To summarize the situation, the fundamental symmetries of

elementary particles seem to point to field theories based on local
internal symmetries connected with structures that can be construct-
ed by using octonions.
The finite octonionic structures are grafted to each point of

space-time, or to each state of an infinite hilbert space in which
the Poincare group or the conformal group act unitarily. The oct-
onionic structure at a point x is related to the one at point x = 0
by an automorphism of this algebraic structure. Hence the para-
meters of the automorphism must be functions of x and we must have

local invariance under the automorphisms of the octonionic struct-
92)
ure. This introduces a connection which describes how the auto-
morphism at point x + dx is obtained from the one at x. As a
result the internal space and the x space become connected in a
fibre bundle structure with the Hilbert space of the external group
as a base and the observables of the charge space or the generators
of the automorphism of their algebra as a fibre. This provides the
geometrical picture associated with a local gauge field theory based
on the automorphism algebra of the internal octonionic observables.
In these lectures we have considered the case in which the

octonionic observables form a Jordan algebra. Then the automorphism
group is an exceptional group F4 or E6 and we are led to a gauge
field theory of quarks and leptons based on exceptional groups.
It is also possible to consider super Jordan algebras for general-

ized Jordan algebras involving both bosonic and fermionic observ-
ab les 1 03) I n t h~s case th e au t omorp h~sm group ~s
. a supergroup. I f
the generalized Jordan algebra is octonionic then its automorphism

is given by an exceptional supergroup. In the case the supergroup
104)
is simple it can be the supergroup G(3) or the supergroup F(4).
The latter has 40 parameters and its Lie subgroup is SO(7) x SU(2)
which admits the phenomenological SU(2) x U(l) x SU(3)c as subgroup.
Here again SO(7) is the norm group of purely imaginary octonions
and SU(3) is the subgroup of the G2 automorphism of the octonionic.
algebra. If the E6 model becomes untenable phenomenologically the
super F(4) is another possible candidate. Instead of leptoquarks
and diquarks it would introduce spin 3/2 gauge particles alongside
with gauge bosons, so it would be similar to supergravity when used
as a local supersymmetry group.
Many problems remain to be solved to complete this general

1138 F.GURSEY
picture. The outstanding ones are concerned with the building of

Fock space (tensor products with exceptional structures), the under-
standing of color confinement and the realization of the ultimate
synthesis with supergravity. Then we might also be able to under-
stand the emergence of a superheavy mass scale that seems necessary
for the unification of strong interactions with electromagnetic and
weak interactions and for the explanation of large renormalization
shifts in experimentally accessible parameters like the Weinberg
angle as well as for the quasi-stability of the proton.
Coming back to Prof. Zichichi' s penetrating questions: "Do we

understand flavor (isospin, strangeness, the new particles)? Should
we believe in quarks, to which we must add: do we understand color?
Do we understand leptons? Can an extension of Quantum Mechanics
lead to a deeper understanding of particle physics?" we may give a
provisional answer to them by saying: There is an underlying quark
structure, leptons and colored quarks do have some kind of symmetry;
there is a good chance that the world is described by a gauge theory
of quarks and leptons and the underlying unifying group or super-
group may be related to octonionic observables. It is of course
possible that new experimental discoveries will steer theory towards
completely different mathematical structures. But a good part of
the newly unveiled structures seems likely to survive.
Acknowledgments
The development of the ideas presented here is due in a large

part to stimulating discussions with my Yale colleagues and with
numerous other physicists and friends including M. Gel I-Mann,
G. Domokos, S. Kovesi-Domokos, M. Gunaydin, P. Ramond, P. Sikivie,
M. Serdaroglu, H. Garland, L. Biedenharn, R. Slansky, L. Michel and
L. Radicati. I would like to express my deepest gratitude to
Prof. zichichi for giving me the opportunity of presenting these

talks and for his extraordinary patience during the time of the
writing up of the lecture notes.
REFERENCES
1. M. Ge11-Mann, CAL-TECH Report CTSL-20 (1961);

Y. Neeman, Nuc1. Phys. ~, 222 (1961);
M. Gell-Mann and Y. Neeman, "The Eightfold Way" (Benjamin,
New York, 1964).
2. M. Gell-Mann, Phys. Lett. ,!!, 214 (1964);
G. Zweig, CERN Report 8182/TH.401 (1964).
3. F. Gursey and L.A. Radicati, Phys. Rev. Lett. 12, 173 (1964);
B. Sakita, Phys. Rev. 136B, 1756 (1964).
4. G. Morpurgo, Physics ~, 95 (1966);
R. Dalitz, 1966 talk reprinted in J.J.J. Kokkedee, The Quark
Model, p. 159 (Benjamin, New York, 1969).
5. D.W. Greenberg, Phys. Rev. Lett. 13, 598 (1964).
6. M.Y. Han and Y. Nambu, Phys. Rev. 139B, 1006 (1965).
7. M. Ge11-Mann, Acta Phys. Austriaca, Supp1. 1, 733 (1972).
8. H. Fritzsch and M. Ge11-Mann, in Proc. 16th Int. Conf. on High
Energy, Chicago 1972, ed. A.D. Jackson and A. Roberts,
(NAL, Batavia, 1973);
W. Bardeen, H. Fritzsch and M. Ge11-Mann, in Scale and Conformal
Symmetry in Hadron Physics, ed. R. Gatto, p. 139 (Wiley,
New York, 1973).
9. S. Weinberg, Phys. Rev. Lett. 11, 1264 (1967).
10. A. Salam, in Elementary Particle Theory, 8th Nobel Symposium,
ed. N. Svartho1m, p. 367 (Stockholm, 1968).
11. P.W. Higgs, Phys. Rev. Lett. 13, 132 (1964);
T.W. Kibble, Phys. Rev. 155, 1554 (1967).
12. C.N. Yang and R.L. Mills, Phys. Rev. ~, 191 (1954).
1140 F. GURSEY
l3. R.P. Feynman, ~n "Proc. 3rd Coni. on High Energy Collisions"

ed. C.N. Yang et al.) p. 237 (New York, Gordon and Breach,
1969).
14. J. Kuti and V. Weisskopf, Phys. Rev. D4, 3418 (1971)
15. H.D. Po1i tzer, Phys. Rev. Lett. 30, l346 (1973);
D.J. Gross and F. Wilczek, Phys. Rev. D8, 3633 (1973) .
16. D.J. Gross and F. Wilczek, Phys. Rev. Lett. 30, l343 (1973);
S. Weinberg, Phys. Rev. Lett. 21, 494 (1973).
17. H. Fritzsch, M. Ge11-Mann and H. Leutwy1er, Phys. Lett. 47B,
365 (1973).
18. D. Amati et a1., Nuovo Cimento 34, 1732 (1964);
J.D. Bjorken and S.L. G1ashow, Phys. Rev. 11, 190 (1964);
Y. Hara, Phys. Rev. 134B, 701 (1964).
19. S.L. G1ashow, J. I1iopou1os and L. Maiani, Phys. Rev. D2,
1285 (1970).
20. J.J. Aubert et a1., Phys. Rev. Lett. 33, 1404 (1974);
J.E. Augustin et a1., Phys. Rev. Lett. 33, 1406 (1974);
G.S. Abrams et a1., Phys. Rev. Lett. 33, 1453 (1974).
21. T. Appe1quist and H.D. Po1itzer, Phys. Rev. Lett. 34,
43 (1975).
22. G. Go1dhaber et a1., Phys. Rev. Lett. lZ, 255 (1976).
23. For a recent review, see M.L. Perl in Proe. of Int. Symp. on
Lepton and Photon Int. at High En., Hamburg 1977,
ed. F. Gutbrod (DESY, 1977).
24. S.W. Herb et a1., Phys. Rev. Lett. 12, 252 (1977).
25. D.Z. Freedman, P. van Nieuwenhuizen and S. Ferrara, Phys.
Rev. D13, 3124 (1976);
s. Deser and B. Zumino, Phys. Rev. Lett. 62B, 335 (1976).
26. J. Wess and B. Zumino, Nue1. Phys. B70, 39 (1974);
D.V. Vo1kov and V.P. Aku1ov, Phys. Lett. 46B, 109 (1973).
27. A. Salam and J. Strathdee, Nuc1. Phys. B87, 85 (1975);
R. Delbourgo, A. Salam and J. Strathdee, Phys. Lett. 51B,

475 (1975).
28, F. Gursey, Nuovo Cimento li, 230 (1960); also in Proc. 196D,
Rochester Conf. p. 572;
M. Gell-Mann and M. Levy, Nuovo Cimento 1&, 705 (1960);
Y. Nambu, Phys. Rev. Lett. ~, 380 (1960);
s. Weinberg, Phys. Rev. Lett. 12, 616 (1966).
29. J. Pati and A. Salam, Phys. Rev. D8, 1240 (1973).
30. F. Gursey, P. Ramond and P. Sikivie, Physics Lett. 60B,
177 (1976).
31. H. Georgiand S. Glashow, Phys. Rev. Lett. 32, 433 (1974).
32. H. Fritzsch and P. Minkowski, Ann. of Physics 93, 222 (1975).
33. J.D. Bjorken and K. Lane, unpublished, see J.D. Bjorken,
Sumnlary Talk in the 1977 Neutrino Conference, (Elbrus, 1977).
Y. Achiman, Phys. Lett. 70B, 187 (1977).
34. G. Segre and H.A. Weldon, Hadronic Journal l, 424 (1978).
35. A.deRujula, H. Georgi and S.L. Glashow, Phys. Rev. D12,
147 (1975).
36. For a review see Prof. Coleman's lectures in these proceedings.
37. For reviews see "Deeper Pathways in High Energy Physics,"
Proc. of Orbis Scientiae 1977, ed. A. Perlmutter and L.F. Scott
(Plenum, New York, 1977).
38. M. Gell-Mann, P. Ramond and J.H. Schwarz, Phys. Lett. 76B,
417 (1978).
39. S. Ferrara, M. Kaku, P.K. Townsend and P. van Nieuwenhuizen,
Nucl. Phys. lli2., 125 (1977).
40. S. MacDowell, Yale preprint (1978).
41. Here is a brief list of some Erice lectures relevant to
SU(6) synunetry:
1964 - G. Zweig: Fractionally Charged Particles and SU(6)
1965 - A. Pais: Higher Symmetries
1142 F. GURSEY
J.S. Bell: Difficulties of Relativistic .SU(6)

L.A. Radicati: The Significance of Internal Symmetries
1966 - L.A. Radicati: Hadron and Lepton Internal Symmetries
M. Ge11-Mann: Relativistic Quark Model and Current
Algebra
B. Zumino: Models of Strong Interactions
1967 - S. Coleman: All possible Symmetries of the S-Matrix
1968 - G. Horpurgo: Quark Model
1969 - W. Thirring: Quark Model
G. Preparata: Relativistic Quark Model
1971 - G. Morpurgo: Vertices in the Quark Model
1972 - Y. Neeman: Hadron Strengths and a System of Hadron
Symmetry
1973 - F. Buccella: From Constituent to Current Quarks
Breaking SU(6)
w
1974 - G. Morpurgo: Composite Particles (of Quarks)
G. Preparata: Massive Quark Model
V. Weisskopf: Quarks in a Bag
1975 - V. Weisskopf: Recent Bag Results
1976 - F. Buccella: Quarks: Color and Octonions
1977 - See Prof. Preparata's lectures in these proceedings.
42. E.P. Wigner, Phys. Rev. 21, 105 (1937).
43. P. Franzini and L.A. Radicati, Phys. Rev. Lett. ~, 322 (1963).
44. A. Arima and F. Iachello, Phys. Rev. Lett. 35, 1069 (1975).
45. C. Moeller and L. Rosenfeld, K. Danske Vidensk. Selsk. Mat.
-fys. Hedd .!Z" 8 (1940).
46. E.C.G. Stueckelberg, Helv. Phys. Acta 11, 347 (1940).
47. W. Rarita and J. Schwinger, Phys. Rev. ~, 436 (1941).
48. J. Schwinger, Phys. Rev. 61, 387 (1942).
49. J .M. Jauch and Ning Hu, Phys. Rev. 65, 289 (1944).
50. S. Coleman, Phys. Rev. l38B, 1262 (1965).
51. S. Weinberg, Phys. Rev. 133B, 1318 (1964), Phys. Rev. 139B,
597 (1965).
52. L.O'Raifeartaigh, Phys. Rev. 139B, 1052 (1965).
53. B. Stech, Zeit. Phys. 144, 214 (1956).
54. H.J. Lipkin and S. Meshkov. Phys. Rev. Lett. ~, 670 (1965).
55. R.P. Feynman, M. Ge11-Mann and G. Zweig, Phys. Rev. Lett. 13,
678 (1964);
R.F. Dashen and M. Ge11-Mann, Phys. Lett. 12, 142, 145 (1965).
56. For a review, see K.G. Wilson, in New Pathways in High Energy
Physics II, ed. A. Perlmutter, p. 243 (Plenum, New York, 1976).
57. J. Me1osh, Phys. Rev. D9, 1095 (1974).
58. P.N. Bogo1iubov, Ann. Institut H. Poincare ~, 163 (1968).
59. A. Chodos, R.L. Jaffe, K. Johnson and V.F. Weisskopf,
Phys. Rev. D9, 3471 (1974).
60. L. Fo1dy and S.A. Wouthuysen, Phys. Rev. ~ 29 (1950).
61. F. Gursey, Phys. Lett. 14, 330 (1965).
62. F.J. Gilman, M. Kugler and S. Meshkov, Phys. Rev. D9,
715 (1974).
63. A.J.C. Hey, J.L. Rosner and J. Weyers, Nuc1. Phys. B61,
205 (1973).
64. F.E. Close, H. Osborn and A.M. Thompson, Nuc1. Phys. B77,
281 (1974).
65. L.H. Chan, Phys. Rev. ~, 2478 (1977).
66. R. Friedberg and T.D. Lee, Phys. Rev. D16, 1096 (1977),
also: "QCD and the Soliton Model of Hadrons" Columbia
preprint CU-TP-118 (1978).
67. J. Rosner, Phys. Reports 11C, 190 (1974);
O.W. Greenberg and C.A. Nelson, Phys. Reports 32C, 69 (1977).
68. J.K. Kim, Phys. Rev. Lett. 1I. 1079 (1967);
C.H. Chan and F.T. Meiere, Phys. Rev. Lett. 11, 299 (1964).
1144 F.GURSEY
69. For a review of the applications of asymptotic freedom, see

H.D. Politzer, Physics Reports l4C, 130 (1974).
70. S. Tani, Prog. Theor. Phys. Kyoto ~, 267 (1951);
See also !1.H.L. Pryce, Proc. Roy. Soc. A195, 62 (1948).
71. M. Cini and B. Touschek, Nuovo Cimento 2, 422 (1958).
72. K.M. Bitar and F. Gursey, Phys. Rev. 164, 1805 (1967).
73. M.M. Nagels, T.A. Rijken and J.J. de Swart, Phys. Rev. D12,
744 (1975), and Phys. Rev. D15, 2547 (1977).
74. For an interesting interpretation of the Zweig rule within
this context, see the Geometro-dynamic Approach to Broken
SU(6) by G. Fogli and G. Preparata, CERN-preprint (1977).
See also Prof. Preparata's lectures in this volume.
75. W. Celmaster, H. Georgi and M. Machacek, Harvard preprint
(1977);
W. Celmaster, Phys. Rev. D15, 1391 (1977).
For specific relativistic models leading to qq bound states
with broken SU(6), see, for example
R.P. Feynman, M. Kislinger and F. Ravndal, Phys. Rev. D3,
2706 (1971);
A. LeYaouanc, L. Oliver, O. Pene and J.C. Raynal, Phys. Rev.
D9, 2636 (1974);
B.H. Kellett, Ann. Phys. ~, 60 (1974), Phys. Rev. DlO,
2269 (1974), Phys. Rev. D15, 3366 (1977);
J.L. Cortes, A. Morales, R. Nunez-Lagos and J. Sanchez-Guillen,
in Proc. VII Gift Seminar on Theor. Phys.,
ed. A. Bramon, p. 195 (1976).
76. H.J. Schnitzer, Phys. Lett. 76B, 461 (1978) and references
therein.
77. For a generalization to SU(6) x SU(3) see F. Buccella,
A. Sciarrino and P. Sorba, Lett. Nuovo Cimento lQ, 455 (1974).
78. S. Meshkov, in Proc. 17th Int. Conf. on High Energy,
ed. J.R. Smith, p. 11-101, (London, 1974).
79. A. J G. Hey, in Proc. 17th Int. ConL on High Energy,

ed. J.R. Smith, p.II-120, (London, 1974).
80. For general discussions of the relation between chira1 symmetry
and SU(6) symmetry, see P. Chang and F. Gursey, Nuovo Cimento,
63A, 617 (1969);
D.G. Ca1di and H. Pagels, Phys. Rev. D14, 809 (1976).
For broken chira1 symmetry see H. Pagels, Phys. Reports 16C,
221 (1975).
81. T. Hagiwara, Y. Kazama and T. Takasugi, Phys. Rev. Lett. 40,
76 (1978);
J.L. Rosner, C. Quigg and H.B. Thacker, Phys. Lett. 74B,
350 (1978).
82. A.J. Buras, J. Ellis, M.K. Gaillard and D.V. Nanopou1os,
Nuc1. Phys. B135, 66 (1978).
83. H. Georgi, H.R. Quinn and S. Heinberg, Phys. Rev. Lett. 12,
451 (1974).
84. T. Appe1quist and J. Carazzone, Phys. Rev. D11, 2856 (1975).
85. F. Gursey and M. Serdarog1u, Lett. Nuovo Cimento ll, 28 (1978).
86. F. Gursey, in 2nd Horkshop on Current Problems in High Energy
Particle Theory, ed. G. Domokos and S. Kovesi-Domokos, p.3,
Johns Hopkins University, Ba1timore,(1978).
87. For a review of the experimental situation, see L. F. Abbott
and R.M. Barnett, SLAC preprint No. 2136, to be published in
Phys. Rev. D (1978).
88. F. Gursey and P. Sikivie, Phys. Rev. Lett. ~, 775 (1976);
P. Ramond, Nuc1. Phys. B110, 214 (1976);
P. Sikivie and F. Gursey, Phys. Rev. D16, 816 (1977).
89. S. G1ashow and S. Heinberg, Phys. Rev. D15, 1958 (1977).
90. D.J. D'Jakonov, Leningrad Physics Institute Report No. 303,
(1977), unpublished.
91. P. Jordan, Nachr. Ges. Hiss. Gottingen, 209 (1933).
1146 F. GURSEY
92. F. Giirsey, "Non-Associative Algebras in Quantum Mechanics and

Particle Physics", Proc. of Conference on Non-Associative
Algebras at the Univ. of Virginia in Charlottesville, 1977,
to be published.
93. P. Jordan, J. von Neumann and E.P. Wigner, Ann. Math. 35,
29 (1934).
94. A.A. Albert, Ann. Math. li, 65 (1934).
95. M. Giinaydin and F. Giirsey, J. Math. Phys. 14, 1651 (1973).
96. H. Freudenthal, Advances in ~~th I, 145 (1965).
97. G.D. Mostow, "Strong Rigidity of Locally Synnnetric Spaces"
(Princeton U. Press, 1973).
98. R. Moufang, Abh. Math. Sem. Univ. Hamburg~, 207 (1933).
99. C. Cheval ley and R.D. Schafer, Proc. Nat'l Acad. Sci. U.S. 36,
137 (1950). The result was anticipated by P. Jordan, Abh. Math.
Sem. Univ. Hamburg 16, 74 (1949).
100. See for example, R.D. Schafer, "An Introduction to Non-
Associative Algebras" (Ac. Press, New York, 1966).
101. M. Giinaydin and F. Giirsey, Phys. Rev. D9, 3387 (1974);
F. Giirsey in "Johns Hopkins Workshop on Current Problems in
High Energy Particle Theory", ed. G. Domokos and S. Kovesi-
Domokos (Baltimore, Md., 1974).
102. J.R. Faulkner, Memoirs of the Am. Math. Soc. No. 104
(Providence, R.I., 1970).
103. F. Gursey and L. Marchildon, J. Math. Phys. 1l, 942 (1977).
104. V. Kac, Comm. Math. Phys. 53, 31, (1977).
DIS C U S S ION
CHAIRMAN: Prof. F. Gursey
Scientific Secretaries: E. Guadagnini, P. Townsend
DISCUSSION No.1
MOTTOLA:
Could you elaborate on the original work of Wigner

and Foldy-Wouthuysen? What is the definition of a Wigner
state and what are the problems with relativistic invari-
ance and positive definiteness of the Hilbert space?
Also, how do you go from a cov~riant to a non-covariant
formalism?
GORSEY:
The Poincare group is non-compact and you want to

represent it on a Hilbert space with positive definite
metric. That is the problem and it was solved by Wigner
in 1939. There are two Casimir invariants of the
Poincare group, - 'P1l1'1'-= M2. and W~ 't.J\I.= Mzs(~ ... 1')(in the
timelike sector). M and s label the states of Hilbert
space. The components of these states are labelled by
the eigenvalues of a complete set of commuting operators.
The set includes M, s, and the three components of ~
The difficulty is to find the last one and that is what
Wigner solved. Physically it is simply one component of
the spin vector, Sz for example, but it is not easy to
express this in terms of the Poincare generators as Sz
is not a covariant quantity. The ket \M1SjPSlZa"> corres-
ponds to wave functions on which the Poincare group must
act unitarily. The norm of this ket is
1147
1148 DISCUSSION
C~3r L <MSj1S~\ t'1$)pSz) = r i p z

J 2s+I J '(~) '2~ I
< 1-lE-JE I)
O~)= J p2
mZ+ 1
which must be invariant. How ~o the Poincare transform-

ations act on these states? d p /El') is Poincare invari-
ant so the unitarity of the transformation requires that
{E. l M,s j ps~) - (E" Ut,'zS2, ~Plo(~'" ') \ MS 'J 1's~ >
...
where U is a (2s+l)x(2s+l) unitary matrix depending on
p and the parameters of the Lorentz transformation, (we
have put P~ = O(tA~ t>" ). The transl~tlons :,,"ct as
'I. rCl + 1. Eo.o
Ta.\M,S'/ = e l M , s !
which is a unitary transformation. Now for each the p
components of 1 \'II, s) ? I ~z) j S2 = (-~ J ' " , S) J
transform like an irreducible representation of the rota-
tion group. Thus the rotation parameters <Xij enter U
as for the usual rotation matrix, i.e. in a ~ independent
way. Only the boost U(~, ~oi ) is p dependent. It has
the form .... -..(~....,.)
... ( E \"'2 1. S. t\)y) b ....')
"Sb ,M, S j f ') =: YE' } e l M ~ p I )
which is a unitary transformation in an infinite dimen-

sional Hilbert space sinc~p is taken into ~and E ~ E'.
S is the spin matrix and t.VtP.,t) are three angles.
i3lV,~) can be found by applying a boost to \ p) to bring
it to rest, making a rotation, and then another boost to
get to Ip'). This transformation is just the Foldy-
Wouthuysen transformation. This is easy to show for
spin~. Expand the positive frequency Dirac field as
~I.t...) r") r:::: ~ Jp

. .,.....,. . .... .... -
el p.x e'l. )',,"1',). (~) C. p
T "x = ~m. -
E
where a.? satisfies o.p=""...a.p so that it cU!~cribes only
the positive frequency components and e\i14 'eX is the
particle wave function ( 1),\ =
~'nh-\tf..I) ') .
E
DISCUSSION 1149
This is not unitary but we can extract a unitary matrix

from it. Since a.t' can be written as t \+~+ ') a. p we have
where
and this is unitary since

E'Z= ~'Z.+ m'Z.
Hence
Hence the Fourier transforms of the non-covariant Wigner

operators Q~ which create the states of Hilbert space are
related to the covariant field operators by the non-local
Foldy-Wouthuysen transformation. If you work out the
coefficients of U(~) on q p and b p you get the 4 wave
functions which you see in all the textbooks, two for
particles and two for antiparticles.
If you use a helicity representation for the states
and choose a direction ?the same methods lead to the
Melosh transformation when p
is taken along the ~ direct-
ion. All of these transformations which take you from
the covariant non-unitary representation of the Lorentz
group to the non-covariant unitary representations are
non local, and can be obtained in standard (after Wigner)
group theoretical ways. This is now called the theory
of induced representations by mathematicians.
JANCEWICZ:
I did not understand the connection between the

Foldy-Wouthuysen transformation and the unification of
symmetries. Can you explain?
1150 DISCUSSION
GURSEY:
J T-+
ap cr-Qp
_13
01 1> (and
5b+p -+ Dr d t"'
Consider a single Dirac particle.
(j'
~
The spin operators
for antiparticles) are
conserved so that they can be considered as internal
degrees of freedom which generate an SU(2) algebra. These
operators !~e not ~ovariant however. The covariant oper-
ators are '\l ~... '\> d x but these are not conserved since
they do no commute with the Dirac Hamiltonian :
II ... -
n="'4m+ l.'("{.V. Under a FOldy-Wouthuysen transformat~on
H _ '1C = lH,,~\)-1 and the covariant spin is replaced by the
non-covariant spins. The non-covariant spins do commute
wi th 1< so they are conserved.
If we have two Dirac particles in a doublet of SU(2)
isospin then the following 15 non-covariant operators
are conserved
JQ+pOi.Q p d p, SQ.tp~iap d'3? j a+r <!i1;j4.\' d'jp
3 I
These generate an SU(4) algebra which unifies SU(2) spin

with SU(2) isospin. This can be verified from the canon-
ical anticommutation relations
{~+f~ ) a.p/~} = &ij ~)O~'- ?') )

which are also not covariant.
TOWNSEND:
What does Coleman's (or O'Raifeartaigh's) theorem

actually state and does it apply to the unification of
other internal symmetries or other space-time symmetries?
GORSEY:
The theorem states that a symmetry group containing

both the Poincare and an internal symmetry group can only
be a symmetry of the S matrix if it is a direct product
of the two. The theorem relies on the non-degeneracy of
the vacuum and on the absence of massless particles so
it does not apply to the conformal group or to spontan-
eously broken symmetries which have massless particles
or a degenerate vacuum. The theorem does not apply to
the physically important SU(2)xSU(2) chiral symmetry
which also involves a massless particle (zero mass pion
in the symmetry limit) and has a degenerate vacuum.
DISCUSSION 1151
Each one particle state is also infinitely degenerate in

that limit. Hence this theorem is too restrictive. Also
supersymmetry which combines internal and external degrees
of freedom and is a symmetry of the S matrix manages to
escape the theorem. SU(6) is an approximate symmetry of
some matrix elements of the S matrix but not a symmetry
of all its elements and this is perfectly consistent with
the no-go theorems.
CARTER:
For a Dirac spinor there are two separately conserv-
ed spins for particles and antiparticles because a Dirac
spinor is not irreducible under the Poincare group. But
for an interacting Dirac particle these spins are not
separately conserved. Is this related to the need to
write particle and antiparticle in an interacting field
theory and what is the status of interacting fields with
respect to the Lorentz and Poincare group?
GORSEY:
The representation theory of the Poincare group as
derived by Wigner is purely for free particles. Starting
from these representations and multiplying them by cer-
tain non-unitary matrices you get the covariant fields.
"Covariant" means "finite dimensional representation of
the Lorentz group" ; the Poincare group has nothing to
do with covariance, its unitary representations are non-
covariant as we have seen. In field theory we also want
the fields to belong to finite, non-unitary and linear
representations of the Lorentz group and this is a diff-
erent requirement.
We impose this condition because we want locality,
necessary for microcausality. The Poincare group does
not act locally on particle states but acts locally on
the fields. This is why, when we reduce a field with
respect to the Poincare group we get all sorts of part-
icles. For example a vector field describes two part-
icles, a 1- and a 0+. We suppress the 0+ with the con-
dition ")",A", = o. For a spinor field we get both part-
icles and anti-particles from the same field.
It is interesting that for the Poincare group Wigner
showed that you can find 3 operators Xi functions of
the Poincare generators, which have the remarkable
1152 DISCUSSION
property that
and are vectors under rotations and odd under parity.

These are just the Heisenberg formulae of quantum
mechanics derived from the representation theory of the
Poincare group. But these X~ are not covariant objects
and transform non-linearly under boosts. The smallest
simple group for which we can get four covariant comm-
uting variables X is the conformal group so that the
representations of~the conformal group can be labelled
in a covariant way. There are three Casimir invariants
J 1 J 2 ,and d (the dimension of the field) and all are
covar1ant.
The state vectors can be labelled as
X.,. ')
3. 1 ;2 , d j ~12..1 ~2'2. , which is sufficient to
describe a state created by a covariant field acting on
the vacuum. So if you want to deduce the space-time
points from group representation theory you have to go
to the conformal group.
The 4 degrees of freedom in the Dirac field nec-
essary for defining a particle spin as well as an anti-
particle spin in SU(6) are related to the requirement of
parity invariance. Otherwise, as for the neutrino, we
can have a two component field which only allows us to
define helicity. For Dirac particles at rest, on the
other hand, the operators .1. (., "6".) separate particle
and antiparticle states and lead to spins ~ l'l",..) S .
For interacting fields these operators can only be
approximately conserved as positive and negative frequ-
encies can no longer be separated by the Foldy-Wouthuysen
transformation.
DISCUSSION No.2 (Scientific Secretaries: A.B. Carter,

M. Mintchev)
MIETTINEN:
My question concerns lepton-quark symmetry in your

model. Usually when one speaks of a symmetry, one imag-
ines that in the zeroth order approximation one has a
single type of fermion f. One then introduces symmetry
breaking, and this "splits" the fermions into various
DISCUSSION 1153
families. Since in particle physics we have a hierarchy

of interactions, we might expect the splitting to take
place along these lines. This is what you have described
this morning, when you separated the fermions first into
quarks and leptons:
Zeroth Order First Order Second Order
/
quarks
-----
----
heavy
medium heavy
light
'"
fermion
----
heavy
leptons ~ medium heavy
light
On the other hand, when one does spectroscopy, one

usually thinks of the splitting as occurring between
groups of particles of similar mass. We then observe
that among the known fermions, the masses organize them-
selves in groups:
(~e) (~) \~) (~) (:,)(:)

, ,
, .A
\ ~
lig'nt medium he.a"j hec!"~
Considering this, one might expect the "splitting" of

fermions to occur along the following lines:
Zeroth Order First Order Second Order
lig~t ----- quarks

leptons
<
/
quarks
fermion med~um heavy
\
leptons
heavy
------ quarks
leptons
Why is this latter approach not the one usually followed?

My question reduces to the following: Is the "up" quark
more closely related to the electron or to the charmed
quark?
1154 DISCUSSION
GORSEY:
It is true that in the context of gauge theories,
the IIsplittingll occurs in two stages which correspond to
the first diagram you have drawn. But this initial
splitting has nothing to do with fermion masses. In the
first stage, we break the symmetry and allow the gauge
bosons to acquire mass. Now it is an experimental fact
that quarks do not transform themselves into leptons
at observable rates. We therefore require that the
gauge boson mass splitting occurs in such a way that the
bosons which transform quarks into leptons be very heavy.
So we have indeed separated quarks and leptons in the
first step, but this separation, is forced upon us by
the stability of the proton and occurs by introducing
asymmetry in the gauge boson sector, not the fermion
sector. In a sense, the symmetry of quarks and leptons
is preserved: they have very similar interactions. They
just do not mix with one another. Also, as I showed, the
quark-lepton symmetry is not necessarily expressed in a
doublet series as you wrote it.
MIETTINEN:
Today we usually leave out gravitation when we begin
to try to unite the interactions. Why do we think it is
better to begin with strong, electromagnetic, and weak
interactions, and save gravity for the future? Einstein
began with gravitation.
GORSEY:
In Einstein's time, only two forces were known:
gravity and electromagnetism. Therefore unification nec-
essarily involved gravitation. Another point concerns
Einstein's philosophy. He wished to do away with the
concept of a force and reduce all dynamics to geometry.
Maybe if Einstein had known about the weak and strong
forces, he would never have approached unification this
way.
KENWAY:
I would like to ask you to comment on the justifi-
cation for considering the unification of non-gravitation-
al interactions within an exceptional group structure.
DISCUSSION 1155
Also, what would be the status of the graviton in such

an approach?
GORSEY:
I have already given you the rationale for the

choice of exceptional groups. You seek a group: 1) with
small rank, so the number of conserved quantum numbers
is the same as the number we observe in nature; 2) whose
smallest representation has a rather large dimension,
so we can put all known leptons in it; 3) from which
color SU(3) emerges naturally. Other criteria are:
4) essential finiteness of color and lavor degrees of
freedom which cannot be extended arbitrarily for except-
ional groups; 5) as I will show tomorrow, exceptional
groups are natural symmetries of extended forms of Quan-
tum Mechanics involving octonions; 6) finally, the
schemes based on E6 and E7 are so far consistent with
experiment.
The only way we know how to unify gravity with the
other interactions is through local supersymmetry. Super-
symmetric Lie algebras have been studied extensively, and
they do not contain exceptional subalgebras larger than
G2, whereas we need at least E6 to incorporate quarks and
leptons.
So from this point of view things do not look good
for a marriage of gravity with other interactions unified
by exceptional groups. But one should note that this
conclusion depends upon the assumption that the general-
ized Jacobi identities for the supersymmetry obey a rel-
ation which is obtained by using complex numbers, i.e.,
an associative algebra. It does not follow that there
do not exist more general algebraic systems which can be
used for supersymmetry and which contain exceptional
groups as subalgebras.
~ICHICHI:
Electrically charged leptons are massive, whereas

neutral leptons are massless. On the other hand, all
quarks have mass. How do you account for this in gauge
theories?
1156 DISCUSSION
GORSEY:
Until we learn how to diagonalize the mass matrix in

such theories, we essentially put the masslessness of the
neutrino in by hand. If you only allow left-handed neut-
rino fields to exist in your multiplets and do not include
a right-handed counterpart, then the neutrino cannot have
a mass. The quarks also differ from leptons in possess-
ing color charge that might be responsible for their mass.
RAJPOOT:
If you do allow a particle to have both right-handed
and left-handed fields, then it is not clear that you can
keep it massless. Even if in the bare fermion mass mat-
rix the field has zero eigenvalue, it can acquire mass
in higher orders through radiative corrections.
GORSEY:
In some models, it is not easy to have particles

which are described only by left-handed fields. It might
be very unnatural to arrange the multiplets so that some
left-handed fields have no right-handed counterparts.
Masslessness of some fields may be guaranteed by a chiral
invariance that survives spontaneous symmetry-breaking.
BUDNY:
If a simple gauge group G unifies the strong, elect-

romagnetic, and weak interactions, there must be at
least two levels of symmetry-breaking: one giving vector
bosons with masses of order 100 GeV and another giving
masses much heavier. Is there some natural mechanism
which gives these two levels?
GORSEY:
Yes. P. Sikivie and I have been studying E7 with

the Higgs bosons in the 912 representation. We use a
quadratic potential and assign vacuum expectation values.
The result of this symmetry breaking is the subgroup
SU(2)xSU(2)xU(1)xSU(3)color , with fifteen degenerate
vector bosons. The rest of the vector bosons you make
very heavy. You can then go on to introduce further
DISCUSSION 1157
breaking, leaving nine of the fifteen degenerate bosons

massless. These are the gluons and the photon.
If M is the mass of a very heavy boson, and m the
mass of a light boson, one-loop calculation gives
...en M
m
= A (.!. _ ~
0( O/s
)
and
\
1+ ~ ex. )
($iY'l2~W ')
otsno~ma"2~c.I
. =A I (
where A, A', B, B' are determined completely by group

theory and the breaking mechanism. Numerically, we get
('1n'...(\2 {)..., ') ~n ~ Yz and O(S ~ 0.25 to this order.
If you put the Higgs bosons in the 1539 represent-
ation, you can further lower SinzGw.
BUNK:
How many free parameters are introduced in the Higgs

sector of your model?
GORSEY:
If we use the Higgs 912-plet there are only two
parameters because one can only construct a single quad-
ratic invariant potential from the 912.
BUNK:
Why do these parameters not show up in "low energy"

considerations? Where do you use these parameters?
GORSEY:
The two parameters correspond to the mass levels of
the vector bosons: they are used to induce this splitt-
ing. Both are very heavy. Their only effects at low
energy are in the renormalization effects that fix ~s
and SiY)z~w.
RAJPOOT:
What about the left-OVer Goldstone bosons in your
model?
1158 DISCUSSION
GORSEY:
These left-over Goldstones are a menace, but we

believe they will pick up masses through radiative corr-
ections. But we have not looked at this problem in
exceptional groups.
RAJPOOT:
19
What is the meaning of the Planck mass, 10 GeV,
which arises in unified gauge theories of strong, weak
and electromagnetic interactions?
GORSEY:
This mass scale arises in connection with the supp-

ression of unwanted quark-lepton transitions. It just
turns out that if you want to maintain the stability of
the proton, at levels compatible with experiment, you
must have vector bosons with mass about equal to the
Planck mass. No inferences need be drawn from this: you
may regard it as a numerical coincidence. On the other
hand one realizes that when you are dealing with masses
this large, you must take gravity into account. The
coincidence is therefore very suggestive.
BACE:
You described the series real numbers, complex numb-

ers, quaternions, octonions and the relation of this
series to the classification of Lie algebras. Why does
the series end with octonions?
GORSEY:
These are the only four so-called "division algebras'!

This means that these are the only quantities for which
a norm can be defined such that the norm of the product
is equal to the product of the norms. This is Hurwitz'
Theorem.
CELMASTER:
Why do we put quarks and leptons into the lowest-

dimensional representation of the gauge group? Is this
DISCUSSION 1159
habit, philosophy, or is it dictated by nature?
G'ORSEY:
If you do not put quarks and leptons into the low-
est representation, you are back to the old problem of the
Eightfold Way. If we put the quarks and leptons into a
larger representation we would be confronted with the
question of what the lowest representation meant. Higher
dimensional representations can always be built from the
lowest-dimensional representation. In the Eightfold Way,
one puts hadrons into higher-dimensional representations
of SU(3) (i.e. octet, decuplet), and the concept of
quarks emerged from wondering what the triplet represent-
ation meant. So you see that if we put quarks and lep-
tons into such a representation, we would have to wonder
whether they were constructed out of objects which be-
longed to the lowest-dimensional representation! The
whole idea of whether quarks and leptons were the funda-
mental fields would be in doubt.
BIGI:
Do you put both fermions and antifermions in the
lowest dimensional representation?
GtJRSEY:
Yes.
BIGI:
Then do you not have fermion number non-conservation
in addition to baryon number non-conservation?
G'ORSEY:
Not at all. That is why I insisted that the gauge
group be complex, so I would have the invariance under
phase transformations with which we characterize ferm-
ionic multiplets. Thus a fermion will not decay into a
boson. Nevertheless, just like baryon number, lepton
number will also be violated in general since some of
the generators of the group that connect leptons to anti-
leptons will have lepton number 2. The associated bosons
1160 DISCUSSION
must be heavy to be compatible with experiment, but not

necessarily superheavy like those which cause baryon
number violation.
CARTER:
You have pointed out that the number of physically

acceptable candidates for the gauge group in unified
theories is finite. There is a finite number of except-
ional groups. The orthogonal and unitary series are
infinite, but the rank of these groups grows explosively
as you ascend the series, whereas nature presents us with
relatively few conserved quantum numbers.
One can therefore foresee a time in which all the
low-rank Lie groups have been investigated and none agrees
with experiment. We would then have to abandon gauge
theories as we know them at present. Are there any other
algebraic structures similar to Lie groups to which we
could turn?
GORSEY:
I believe that we must find a low-lying group which

describes nature or look beyond finite groups. I just
cannot believe that SU(79) or SU(I03) is the invariance
group of nature. The number of quarks and leptons is
either small or infinite. If we keep finding quarks and
leptons for too long, I will begin to look beyond finite
groups, to infinite groups which induce transformations
among an infinite number of quarks and leptons.
LIPKIN:
For a pedestrian description of the theorem showing

that space-time combined with internal symmetries leads
to a trivial S-matrix, consider a scattering experiment
with a neutron beam on a proton target. Lorentz invar-
iance says that the S-matrix would not be changed by
putting the beam and target on a space ship and invariance
under a group which combines isospin with the Lorentz
Group in a non-trivial way would require the S-matrix not
to change when the neutron beam and proton target are put
on two different space ships moving in different direct-
ions. This is possible only if there is no n-p scattering
DISCUSSION 1161
GORSEY:
This is an illustration of the fact that SU(6) is
not an invariance of the S-matrix.
DISCUSSION No.3 (Scientific Secretaries: F.J. De Luccia,

E. Guadagnini)
WIGHTMAN:
How is your formalism of states and observables in

terms of Jordan algebras related to field theory?
GORSEY:
The Jordan algebras describe the internal states.

To obtain the full structure of field theory one attaches
a Jordan algebra to each point of space-time. Fields are
sections of the resulting bundle.
HINCHLIFFE:
Can you comment on the status of supergravity and in
particular on the status of exceptional groups in attempts
to unify gravity with other interactions?
GORSEY:
I am not an expert on supergravity or supersymmetry.
What I know is what a layman knows, among theorists. All
the simple Lie superalgebras have been classified by Kac,
Kaplansky, and Freund by assuming two axioms : 1) You
have anticommutators as well as commutators, so you have
a generalized Lie algebra, a graded Lie algebra. 2) You
have something which corresponds to the Jacobi identity,
and this is obtained by taking the usual Jacobi identity
and putting in some anticommuting numbers, assuming that
these anticommuting numbers are associative. Then you
get a generalized form of the Jacobi identity which will
hold for generators which are anticommuting. Once you
have these two axioms you can deduce all systems which
satisfy them, and these have been classified. You also
want simplicity, so you look for the simple systems.
You generalize the Killing metric, and demand that it be
non-singular. This then gives you all the simple algebras.
1162 DISCUSSION
Then you can look at these algebras as extensions of

their Lie subalgebras. Their Lie subalgebras are not
simple, in general. They are semisimple, but the whole
thing is simple when extended with the new Fermi-like
generators. You find that the only exceptional group
that appears in the Lie subalgebra is G2 . But the
groups which are characteristic of the exceptional quan-
tum mechanical spaces do not appear. Now these could
appear in a more general system when you assume that the
anticommuting numbers are not associative. I can give
you a very simple example. The octonionic units I de-
fined this morning satisfy the relations :
'Lti U ) + U.1 u.\ =0 ) -i.;l:j , i J ,1 = '1,2.)3,
and these form an anticommuting system, but they are not
associative. So if you put these in the Jacobi identity,
you get something different from what is postulated by
Kac. There may well be more general algebras which start
from a still more general Jacobi identity, and there may
well be other exceptional groups appearing in these.
This is still an open problem. So that's the super-
symmetry part.
You also asked about supergravity. The supersymmetr-
ies are used as symmetries of bosons and fermions, and
the one which occurs in supergravity is a system which
has a symmetry between the spin-2 graviton and a new
particle which has spin 3/2. That gives a generalization
of Einstein's theory. Once you have that, you can try
to put in more fields, which will have internal degrees
of freedom, for example, spin ~ and spin 0, and see if
you get a system with two spin-3/2 fields and two spin-~
fields, so you will get as an internal symmetry U(l) or
SU(2) in more general cases. These have been constructed
by many groups. If you go on this way, it has been shown
by Gell-Mann and Ne'eman that in principle you could go
up to SU(8), but no further if you assume that the grav-
iton is a singlet. However, more recently, van Nieuwen-
huizen and his collaborators have generalized Weyl's
theory, which is scale invariant, in a supersymmetric
way, instead of Einstein's theory. They find that in this
way one can have SU(N) as an internal symmetry group, and
this is much nicer because it allows more freedom. There
may be more general models which will allow the other Lie
groups to come in, SU(N), Sp(N), up to some N, and which
will correspond to some classical superalgebras, and
there may even be some exotic ones if you try to general-
ize the Jacobi identity. So the end is not in sight, and
DISCUSSION 1163
these models hold the promise of unifying gravity and

internal symmetri~s in a formal way.
The models that have already been constructed have
one characteristic feature: If you leave ou~ gravity
and crank up supersymmetric field theories, then they
have a very nice property - they are renormalizable.
Once you put in gravity you get systems which are not
renormalizable. But in the case of supergravity some-
thing very interesting happens. Although it is not
renormalizable at the one-loop and two loop levels, it is
finite to these orders, so the theory is even better
behaved than renormalizable field theories. No one knows
what happens at the three-loop level, where divergences
may occur. But if one constructs systems that admit
maximum symmetry, they may also exhibit maximum smooth-
ness in their short distance behavior. We saw another
example of the same phenomenon in my first lecture on
the Moller-Rosenfeld model. You put more symmetry in
the Lagrangian and you get a well behaved wave function.
In gauge theories the same thing happens. You inject
local symmetries (gauge invariance> and you get Yang-
Mills theories, which are renormalizable. So maybe in
the end we will get a finite unified theory which will
also incorporate internal symmetries. Now are such models
physical or not? Only time will tell, when we can work
out their experimental consequences and confront them
with experiment. Although there is not much we can say
at the moment, all the same, the program that links the
regularity of solutions with maximum symmetry remains an
exciting prospect, and that is what is so stimulating
about supersymmetry, although it has not led to any phy-
sical scheme up to now.
CARTER:
You have said that if one investigates supersymmetry
algebras derived from generalized Jacobi identities which
are in turn derived using non-associative algebras, i.e.,
octon-ions, then you might find that these supersymmetry
algebras have exceptional subalgebras. Then we could
unite gravity with an elementary particle theory based on
an exceptional group. That would be nice, but would the
fact that you used octonions to derive the generalized
Jacobi identities mean that you had to do all the physics
with octonions?
1164 DISCUSSION
Gt'JRSEY:
It would be very nice indeed, and octonions would

playa fundamental role in physics. But no one has yet
shown that one can construct supersymmetric algebras
which have exceptional subalgebras, so we do not know
whether such a situation will occur.
MOTTOLA:
We know that real numbers do quite well in classical

physics. You can describe everything in terms of
Newton's laws and electrodynamics until you get down to
distances like the size of atoms, and that is how complex
numbers obtain their usefulness. Is there some fundament-
al physical constant, analogous to Planck's constant,
which tells us when octonions become important?
GURSEY:
No, as far as I can see, there is no fundamental

constant which comes in. You would like a fundamental
constant such that when that goes to zero the non-associat-
ivity disappears. I don't see how this can come in.
MIETTINEN:
I have a general question. Many people feel that

the time to really find the group theoretical structure
of quarks and leptons is at hand. Could you make a pre-
diction? Will you or someone else in the next few years
find the group theoretical structure of quarks and
leptons?
GORSEY:
I don't want to make any prophecies. The fact that

I have worked on these topics for a number of years shows
that I must believe that there is a grain of truth in
them. But to answer your question, I would not like to
commit myself. If the octonionic quantum mechanics that
has been around for more than forty years turns out to
give a correct description of the internal symmetries in
particle physics, perhaps this would be too good to be
true.
THE BEST WHY
A. Sanda
The Rockefeller University, Department of Physics
NEW YORK, N.Y. 10021, USA
Do we understand J/~ production?
Why two facts presented at this school cannot be understood in the

simplest way ?
Does J/~ production hold the key to understand gluons ?
Fact 1
We have learned from M. Chen that
'1/~
r ? t> p ?
(a) (b) (c)
(a) accounts for a major fraction of J production in pp -+ J + X;

(b) leads to pp -+ J + ~ + X not seen experimentally, and
1165
1166 A. SANDA
(c) leads to
crlpj> ... ~.. )() c: 1

cr(pp+:T+)( )
Let us take (a) as our model for J production. Then extending the
simple model,
:rN ~: q
can be computed from the hadronic decay of J/~:
Once this is known
can be related to
We obtain
('=
THE BEST WHY 1167
The details of this calculation are as follows:
We. want to compute the ratio of
and
First get
from the decay of J:
q. 'Z
-
~I~I =
21~r
\~:r I~r
Using this ratio
1168 A. SANDA
Narrow with approximation
('"-
2M~ d::M Z (1)~-+e"'e--+x) 2m-;r
~
(~~)
2.
;
=
~ S eN-
ee d~dM2
dM~(p~~J+)() _1 S
See
C~)~
of; ltm.,-~ 1
-~ee
.f'a
-L = 11
.of1\"
we thus get
r= - j
THE BEST WHY 1169
Fact 2
Now the experimental value for r was presented by Smith:
~e"h -- 1 ;Co 1
".... 200
What did we do wrong? Is J/~ production more complicated? Is the

gluon contribution
p
important? (No, because this would give (pp pp) for J production.)
Do we understand the IZO rule?
STATUS OF THE SUBNUCLEAR WHYs
Antonino zichichi
CERN
Geneva, Switzerland
INTRODUCTION
The purpose of this lecture is to summarize what we have learned

at this School. The order will be that of the official program, as
presented in the Opening Lecture.
STATUS OF THE WHYs
1. Do We Understand Isospin?
2. Do We Understand Strangeness?
The only contribution in the field of strange particles is the

result obtained by the ABCLV Collaboration 2 ) for the spin of the n-,
shown in Fig. 2.1. As you know, the n- is expected to be the tenth
member of the (3/2)+ decuplet. The mass obtained from the equal
spacing rule, the strangeness (-3), and the isospin value (I = 0)
are the expected ones. So far nothing was known for its spin: it
should be ~2. The results reported at this School show that the
(spin 12) value can be excluded on the basis of the fact that it
1171
A. ZICHICHI
1172
_d
rLcms
if)
20
f-
+r---t
Z
w
>
w
0
w 10
f-
I
t++t
<.:)
w
3:
0
-I 0
COS EJ *
Fig. 2.1 Experimental results on the Q-
spin measurement from the ABCLV Collaboration
STATUS OF THE SUBNUCLEAR WHYS 1173
corresponds to a confidence level below 0.1%. The great news here

would have been to find that the sr spin was not %.
3. Do We Understand the New Particles?
The most spectacular news is the results from Fermilab

(Lederman and collaborators). The inclusive study of (~+~-) pai rs 3)
produced by 400 GeV protons in the reaction
+ - .
p + nucleus + ~ ~ + anythlng
has produced a new bump, called T, at H(~+~-) '" 9.5 GeV with
FWHU '" (1.16 0.09) GeV and a cross-section
~01 BR = (3.4 0.3) x 10- 37 cm 2 /GeV/nucleus

y y=o
The interpretation of this bump lS in terms of the "wanted" new

quarks, anticipated at the time of the anomalous behaviour of high-
energy neutrino reactions (y anomaly, Wh threshold, 0 V/0V rising
versus Ev) and needed in order to explain CP violation in terms of
standard weak interaction coupling among all basic quarks (see
Section 4). As we will see in Section 7, none of the high-energy
neutrino effects are confirmed by the WAl-CERN experiment.
The "top" and "bottom" quarks would have disappeared were it

not for this new and great discovery. Let me remark that this bump
should consist of at least three bumps, according to the theoreti-
cal predictions of Gottfried and Eichten. In fact, if other flavours
exist in addition to charm, and if a new flavour is associated with
a quark mass of '\, 5 GeV, the lowest "open-new-flavour" state has a
mass higher than the third "hidden-new-flavour" state. In the
"charm" case the threshold for the open-charm state (D meson) is
above the second hidden-charm state ~'. A more detailed analysis
of this new effect seems to agree with these theoretical predic-
tions. In fact, it should be noticed that the "single-particle fit"
1174 A. ZICHICHI
has a poor X2 , and this 1S a serious hint in favour of more than

one state in the bump. The Lederman bump would therefore corres-
pond to the simultaneous discovery of three new particles.
Another interesting result 4 ) is the "scaling" in JI1/! produc-

tion (Fig. 3.1). Unfortunately there are only two data points on
l}J' production. As we will see in Section 6 (Fig. 6.2) the "con-
tinuum" of prompt "lepton pair" production scales identically4).
It is therefore interesting to compare the data. To get dimensions
of J/l}J production the same as "continuum", multiply by M2; also
consider do/dy, not B(do/dy):
[M3 (d 2o /dydH) Jcontinuum
1
[M2 (do/dy) yl1/! '" 200
Conclusion: (J/l}J) and continuum scale in the same way. Only the
"coupling" strength differs by a factor of '" 200. Notice, however,
that to the J/l}J inclusive production there contribute other particles
decaying into (J/l}J). So the factor (200) is certainly the upper
limit for the "coupling" which produces (JI1/!) with respect to the
continuum. Repeating the same comparison for l}J' production we get:
[M 2 (dO/dy)]JI1/!
'" 6
[M2(dO/dy)]l}J1
Finally, let me mention a result obtained at CERN 1n the Omega

spectrometerS) by the CMNER Collaboration:
r = ..;.o-;(;.:J:--!p:..;r=-o=-d:;u=c-=t.:;:i-=o..;.n:.....:f~r=-o:...;m:::....p"-=-r..;.o.:;:t..;.o..:;n;.::s...::)_~ 0.15 0.08

o(J production from antiprotons)
This result rules out a quark-charmed fusion model and a gluon

fusion model, both predicting r '" 1. A mixture of several produc-
tion mechanisms is needed in order to explain these data.
10 31

t I [aUncertainty
C0 bb eo. 66"1. scal1
o Snyder et 01.
D Bronson et 01.
1032 A Busser et 01 .
Nagy et 01.
... Antipov et 01.
x Aubert et 01.
N
E 16 33
0
0
">- "
"t"t, IjJ'
bl>-
"0 "0
CD 1634
'y
/\
16 35
B dCTI =7.511034el47MV\ ,
dy y=o
1036
a .1 .2 .7
Fig. 3.1 Dependence on beam energy: B da/dy for J/~ and ~'
production in hadron collisions
1176 A. ZICHICHI
Let us now go to the (e+e-) annihilation process for the pro-

duction of new particles 6 ). Start with the hidden charm states.
The status of the charmonium levels is shown in Fig. 3.2. The
J PC 1+- state is still missing. As you know, it is expected to
be in the mass range (3,415 to 3,550) MeV unless it is pushed down
by a large triplet-singlet shift.
The open-charm states of the 0- l6-plet of SU(4), based on the

quark flavours u, d, c, s, is by now complete. Figure 3.3 shows
the structure of the pseudoscalar SU(4) multiplet. The quark com-
position of the various states is also shown. Notice that the
structure of the vector 1- l6-plet is analogous in terms of the
quark's structure. The only difference is the spin alignment of
the (qq) pair: antiparallel for the 0- and parallel for the 1-
multiplets, the angular momentum for the (qq) being always zero.
Table 3.1 shows the summary of the open-charm meson masses

(pseudoscalar and vector states) as reported at this School. Notice
the equality, within errors, of the mass difference between the
vector and the pseudoscalar states for the D and F mesons. Notice
also that M1-* is determined ~n two independent ways 7): i) measuring
MF* and MF ; ii) from F* + F + y.
The other relevant points are:
i) the absence of molecular charmonium;
ii) the absence of charm-burning mechanism. In fact the inclusive

J-production is missing at 4.028 GeV.
Concerning the weak decays of charmed mesons:
i) they are compatible with (V-A) and the Cabibbo angle;
ii) there is no (20) enhancement. Let me elaborate on this point.

Notice that for the (u, d, s) flavours there is the famous
octet enhancement. The introduction of a new heavy quark, c,
does not seem to extend this enhancement in the higher symmetry
structure. In the decay of charmed particles the truth is not
'" (4415) _ CHARMED HADRONS

4.4
4.2
",(4100)
-------- ..... CHARMED HADRONS
4.0
THRESHOLD FOR
\ OPEN CHARM
3.8 \ 3.73 GeV '" (3772) n 3 01
._.- .. ----_._-_._._-_.-._ .. -.--_._._._._._-_._.
'" (3685)
>
~ 3.6 ____Lo_____
'P? X (3555)
CI)
X (3510)
CI)
X(3455)
~
2 1S 0 1 \ X (3415)
1 \
~ 3.4 I \
\
,
I
1 \Y 11'11'
t
HADRONS?
\
\
\
,
."
3.2 \\
3.0
Y
HADRONS
1 's0 11'+ 11'- ~411'
2.8
K+ K- K+ K-+211'
2.6 HADRONS? 2y
o 2,1,0
SPIN
Fig. 3.2 Charmonium bound states

1178 A. ZICHICHI
;c
I
I
I
I
I
Fig. 3.3 The pseudoscalar SU(4) multiplet

Table 3.1
Summary of charmed meson masses
SPEAR-LBL
DO 1863.6 2 MeV
D+ 1868.7 2 MeV
(D*)o 2006 1.5 HeV
(D*)+ 2008 3.0 MeV
(D*)+ _ DO 145.3 0.5 MeV
8 5.1 2.8 } +
t1 + 8 HO
8* 2.6 2.5
DASP
F+ 2030 60 MeV
(F*)+ 2140 60 MeV

(F*)+ - F+ 120 40 MeV
too far from the naive approach: counting the degrees of free-
dom. The virtual weak boson emitted by the decaying charmed
quark can either go into a lepton pair, W+ + (v e+) and
e
(~~ ~+), or into a quark pair, W+ + (du)red' (du)yellow'
(du)blue. Therefore we have 2 degrees of freedom for le?tonic
final states and 3 for hadronic ones (because of colour). This
implies that 40% of all decays should be semileptonic and 60%
hadronic. These naive quark-counting rules do not work for
the decay of strange particles. For example (A + S)/(A + all) ~
~ 10- 3 , because the non-leptonic decays of the A are enhanced

in the ILH I = liz channel. Only if the I"'I I = % decay is for-
bidden, as in the charged kaon case, does the semileptonic
1180 A. ZICHICHI
rate compare well with the hadronic one. If an enhancement

like ~I = 12 was at work in the charm case, we would have ex-
pected a large suppression of the semileptonic rate. The rate
of electron-hadron events observed in (e+e-) annihilation at
DESY implies a large semileptonic branching ratio for charmed
particles, well compatible with the naive quark counting rule
mentioned above. In particular, as reported by Wiik7) for the
pseudoscalar D meson:
(D + evX) _ +
(D + all) - (11 - 3)%
So, adding the D(~VX) channel, we reach a level of ~ 20%, which

is very far from numbers such as 10- 3 It seems that the non-
leptonic decays of charmed particles are free of all those
conditions such as the ~I = 12 rule that we have in the case
of strange particles. This is not so unexpected. Remember
that the enhancement of a given representation, for example,
the octet for SU(3), is understood in terms of the fact that
at short distances the theory is more symmetric than in general.
As is well known, SU(3) is based on three quasi-mass degenerate
quarks, and the predictions of the octet enhancement are well
satisfied experimentally. For example, the Gell-Mann- Lee-
Sugawara relation for S-wave amplitudes,
2S(~- + ATI-) = I3S(E+ + pTIo) + S(A + pTI-)

+
~~------------~--
+ ________-J'
(4.08 0.04) (4.04 0.05)
is satisfied at the 10% level. The natural extension of the

SU(3) octet enhancement is a 20 in SU(4). In this case, one
additional relation is obtained for S-wave amplitudes, the
Altarelli-Cabibbo-l1aiani condition:
1 S(E++pTI o) ,
73
+
1.48 0.85
which is clearly violated. This means that the addition of

the fourth quark (c), whose mass is far from being degenerate
with the masses of the other three (u, d, s), has stopped the
natural development of a symmetry. This "symmetry stop" happens
in spite of the fact that, according to Wilson 8 ), the enhance-
ment mechanism should arise from short-distance effects
(A ~ l/~), where MW is the weak intermediate boson mass; and
this is notably much greater than the fourth quark mass M
c
Finally" let me close the "new particles" section with the
study of possible existence of exotic states. The only hope here
is the relatively narrow spikes of Figs. 3.4 and 3.5 reported by
Nilsson 9 ). When the existence of these spikes is definitely esta-
blished, a new series of particle states will be started. The
great game here is to see if new multiquark structures really exist.
These structures can be obtained if nature allows not only three

quarks in a baryon (qqq) but also a quark-antiquark in a ~eson (qq)
to stay together. For example, take a state with two quarks and
two antiquarks. These four objects can be arranged inside a hadron
in two different ways: either by pairing two quarks (qq) and two
antiquarks (qq), or by pairing a quark-antiquark (qq) twice. These
states are called baryonium and mesonium, respectively. There are
other multiquark states, with baryon number 0, 1, 2, and different
quarks grouping to form various hadronic states, as shown in
Table 3.2.
Other possible states are for example (qqqqq); this is ob-

tained if in an antibaryon configuration we replace two of the
antiquarks by diquarks:
(qqq)
\t,---~\,
(qq) (qq)
1182 A. ZICHICHI
2020! 3 MeV
30 ~
N
,---
2204!5 MeV
~
~
~ 20
0 t935!3 MeV
C\J
"- ~
if)
f-
Z
w 10
>
w
OL-~~~I~~__~____~~==__
2000
pp invariant mass
Fig. 3.4 pp mass spectrum for TI-P + (PFTI-) pp with PFTI

in the 6(1230) band and cos e pp < 0 for the pp decay
angle
Total unweighted pp,.".- mass spectrum at 16 GeV/c
!l3998 events)
300
250
, 2.15t01 GeV
~
Q,
200
II>
C
-...
Q)
>
Q)
o 150
Q)
.J:l
E
::I
Z
100
50
O~ ____~______~______~____~________~
2.0 2.5 3.0 3.5 4.0
Fig. 3.5 Mass of pp n- ln 16 GeV/c n-p + (ppn-)p reactions

1184 Ao ZICHICHI
Table 3.2
New multiquark hadronic states
- quark o - antiquark B = baryon number
( : g)
baryonium ,
I
C_....:...~_--.::~=----)
}
B o
mesonium
( ~ eee )
mesobaryon } B
1
pseudobaryon
c""----e_e_e___
e..:..e_e__) d 1Ob aryon 1st} B
2
( : ::) dibaryon
2nd
Concerning their decay rates, there are several mechanisms:
i) dissociation:
C: 0
0 ) .- ~
ii) simple cutting: ~
(: 0
0 ) C(V 8)
,
iii) rearrangements of (qq):
c: g)
(;) (;)
(: ~g) -----'"
iv) dissociation:
~
:)
~~
These mechanisms correspond to different decay rates: for
example, the "dissociation" is super-allowed. The exotic states
can decay into either a meson-meson or a baryon-meson system. Some
of them are expected to be narrow, others should be relatively
broad. If all these speculations are valid, a large set of new
states should show up.
4. Do We Understand Symmetry Breaking?
The status of our knowledge remains unchanged for
(1 + As) ~n the semistrong interactions which break SU(3),
(1 + T 3) in the electromagnetic interactions which break STJ(2), and
(1 + 1'5) ~n the weak interactions which violate P, C.
The new development refers to CP or T violation in the super-

weak interactions. These can be incorporated in the structure of
the weak interactions if at least six quarks are present 10 ) as the
basis of the hadronic world: u, d, c, s, t, b.
In order to understand T violation it is necessary to extend

the original Cabibbo mixing between down (d) and strange (s) quarks
as shown in Fig. 4.1. Three mixing angles and a phase are needed.
The present knowledge of these angles is as follows:
sin 2 8] = 0.05 -- this is the original Cabibbo angle;

sin 2 8 2 must be small and is at present compatible with zero, as
can be deduced from the study of the process V + d + C + W,
W
which is proportional to sin 2 81 cos 2 82
A. ZICHICHI
1186
u c
I
d
1s
(0 )
~\
ds b
Fig. 4.1 SU(2) quark structure a) before and b) after the

generalized Cabibbo mixing.
sin 2 6 3 cannot be large because this would lead to violation of

weak interaction universality. In fact the old belief that
(u ++ s) transitions are governed by sin 2 6 1 is now re-
placed by sin 2 6 1 cos 2 6 3 The Cabibbo universality ~s
therefore controlled by the extra factor cos 2 6 3
For the sake of clarity, in Fig. 4.2 we show how the various mixing
angles come into the different quark transitions.
Notice that ~n this theory with six quarks, three mixing angles,
and a phase, the CP violating amplitude is in general depending on
the product
If we take all angles equal to 6 1 ~ 6Cabibbo' we have a CP violating

amplitude not so far from the experimental value.
The physical significance of these mixings is as follows. The

structure of the quarks, before mixing, is given by the three SU(2)
doublets:
The only allowed transitions are u ++ d, c ++ s, t ++ b. Without

mixing, this is the end of the story. Owing to the mixing mecha-
nism among the (-13) charge degenerate quarks (d, s, b), it is
possible to have transitions (u ++ d, s, b), (c ++ d, s, b),
(t ++ d, s, b). Notice that even after the generalized "Cabibbo"
mixing, transitions among charge-degenerate quarks are still for-
bidden. It should be emphasized that equivalent results would be
obtained if the mixing is described between the (+13) charge-
degenerate quarks (u, c, t). As mentioned above, this mixing is a
rotation in SU(2) of a complex nature; it needs three angles and
one phase, i.e. four parameters.
1188 A. ZICHICHI
Fig. 4.2 Showing how the various m1x1ng angles come

into the different quark transitions
s. sin 8.
1 1
c. cos 8.
1 1
Do we have any reason to expect mixing between charge-degeneratE

quark states? The answer is Yes. Recall the gauge symmetry of
weak and electromagnetic interactions, SU(2) x U(l) 11) This
gauge symmetry is needed in order to have a renormalizable theory.
But it is not enough. All particles of the theory should be mass-
less and all interactions invariant under a symmetry transformation.
Nature is made with masses: quarks are massive. Quark masses
break weak isospin because mass terms involve left and right
states. But L-states are isospinors, while R-states are isoscalars
(see Section 5). Renormalizability remains, even if the symmetry
is broken, provided that the breaking goes through coupling to
Higgs scalars I2 ) . Since symmetry breaking and Higgs coupling con-
serve charge, the mass matrices generated by the Higgs mechanism
mix only charge-degenerate fields. This is exactly what lS ob-
served in weak interactions. So the physical meaning of CP viola-
tion is to tell us that at least SlX quarks exist in nature and
that they mix in such a way as to keep the weak interactions re-
normalizable. And the original Cabibbo discovery was the first
evidence for this new understanding of weak interaction phenomena.
Without generalized Cabibbo mixing, the strange quark could not go
into up quarks but only into charm. In other words, c ~ s tran-
sition would be the only one allowed. Think of the consequences
for our standard physics program, before the discovery of charm, if
this was true.
Not all symmetries of nature are broken. In the unbroken

front we have the colour symmetry group SU(3) 13), which masters
c
the interaction between spin 12 coloured-flavoured quarks with
vector-coloured-flavourless gluons. This SU(3) should be the
c
gauge group of strong interactions, and this group must commute
with the gauge group of weak interactions, in order to have a re-
normalizable theory of these interactions, all together.
1190 A. ZICHICHI
5. Do We Understand Leptons?
The present status of leptons is
(5.1)
The best evidence for the new lepton T comes from (e~) final states
originally observed at SPEAR 14 ) and later at DORIS. The latest re-
sUlts 15 ) for the mass value are from two groups at DORIS:
Heidelberg-DESY group 1. 787 0.010 GeV

0.018
DASP group M(T) = 1.807 0.020 GeV.
These values are obtained assuming (V-A) for the T decay.
Another recent result from DORIS 1 5) is the value of the

branching ratio
T -)- IT\!
T
(9.00 2.9)%
T -)- all
Let me recall that the search for heavy leptons, exactly along the
line which has been found to be the cleanest, i.e. (~e) pairs in
the final state, has been developed originally in Frascati with
ADONE 16 ) . The machine energy was a few hundred HeV below threshold,
and the only result obtained on the (e~) pairs was a limit 17 ) on
the heavy lepton mass: Mh ~ 1 GeV.
The key point is, of course, to establish if the T has its

own neutrino \!T' as would be suggested by the symmetry group of
weak and e.m. interactions SU(2) x U(l), whose doublet structure
for leptons is shown above [formula (S.l)J. For clarity, we show
w
~n Table 5.1 the weak quantum numbers, T = weak isospin,
SW = weak strangeness, NW = particle number, corresponding to all
leptons and quarks known or seriously suspected so far. Notice
that the electric charge Qe.m. is given by a Gell-Hann - Nishij ima-like
relation:
Nw + SW
Qe m. = TW + ':":---..",.---'-_
3 2
All leptons (left- and right-handed) have "particle number" NW -l.

The L-handed are isospinors, and have TW = 12 and SW = O. The
R-handed are isoscalars, TW = 0, but have weaf strangeness, SW = +1
for (V e , V~, VT )*), and SW = -1 for (e-, ~-, T-). All quarks (left-
and right-handed) have NW = 13. The L-quarks are isospinors, TW = 12,
but have SW = O. The R-quarks are isoscalars TW = 0, but have SW = +1
for (u, c, t) and SW = -1 for (dC' sC' bC); these states are "Cabibbo
mixed states".
Table 5.1
Values of intrinsic weak quantum numbers for left- and
right-handed leptons and quarks in the SU(2) x U(l)
Salam-Weinberg gauge theory of weak and electromagnetic
interactions
r:
(: t (;t
NW SW YN Qe.m.
1
- 1 +- 0 - 1 0
2
CIl
Z
(::t - 1 - -21 0 - 1 - 1
0
E-<
------------------------ ----- --- - ---- ----- ----
p.,
ril (e-)R (!n R (-r-)R - 1 0 - 1 - 2 - 1
....:I
No neutrinos (R) because m,,= all zero
(:1 (.J CJ
Ul 1 1 1 2
+- 0 +- +-
~ 3 2 3 3
1
1 1
- -31
0
u - 0 +-
3 -2 3
M
X --------------------------- r----- ----- ----- ----- -----
CIl
2
::.::
p:: (dC)R (sC)R (bC)R -
1
3
0 - 1 - 3 -"31
~
~ 1 4 2
G (u)R (c)R (t)R 0 + 1 - +-
"3 3 3
*) If neutrinos are massless, these states do not exist.

1192 A. ZICHICHI
Notice that we have SlX leptons and SlX quarks, i.e. an equal
number of quarks and leptons. Why is this needed? It is needed
in order to avoid the Adler-Bell-Jackiw anomaly18) of the axial
vector currents, which spoils renormalizability. This anomaly
comes from diagrams of the type
where W3 lS the weak gauge boson coupled to the weak isospin (T~),
and B lS the weak gauge boson coupled to the weak hypercharge yW.
The arrowed lines indicate the basic fermions of the theory (lep-
tons and quarks). The basic Lagrangian (Salam-Weinberg)ll) is in
fact
!.t
e.m.+weak g[L: ~t ~ wt) . w+ g'(Li ~iywWi) B ,
i
where T is the weak isospin operator and yW the weak hyper charge
operator. The ratio of the two coupling constants
tan 8 g' / g
w
is the famous weak angle 8. Notice that this angle is nothing more
w
than the measurement of the mixing between the two symmetry groups
which put leptons and quarks on the same footing, SU(2) and U(l).
This mixing should obviously be the same for quarks and leptons.
In order to cancel the ABJ anomaly, the electric charges of all
leptons and quarks should be zero:
So, given six leptons we ought to have six quarks with the correct
electric charge states. If the electric charge of the quarks is
fractional, as in the standard Gell-Mann- Zweig model, three colours
are needed in order to get zero for the total charges of all lep-
tons plus quarks. In fact, with (e-, ~-, T-), the total lepton
charge is -3, while the total quark charge with (u d), (c s), (t b)
is (%-%) + (213-%) + (%-%) = +1, which becomes +3 if we intro-
duce three colours. The introduction of colour is essential if we
are to avoid the above-mentioned triangle anomalies.
We believe that this doublet lepton and quark structure ~s a

very simple way of unifying weak and electromagnetic interactions
of leptons and quarks. However, the experimental results of heavy-
lepton physics need further detailed work to establish that we are
really dealing with a new lepton with its own neutrino, as we would
like the case to be. This is why it is important to exclude other
possibilities, for example the so-called "gauge leptons", namely
those heavy leptons invented in order to cancel divergences in weak
interaction processes before the discovery of neutral currents.
These would be electron- and muon-like leptons E+, l~, which would
make up triplets such as (e-v E+), (~-v M+); all particles in a
e ~
bracket having the same leptonic number, either electron-like for
the first triplet, or muon-like for the second.
This is not the end of the story. Suppose (we are physicists,
~.e. the facts are needed to establish the truth) that the process
coming into play is
and suppose that
+ V
e
+ V
~
1194 A. ZICHICHI
In the final state we still have patterns analogous to those seen

so far, because the charge 13 quark would not have been detected,
thus simulating a missing neutrino. The spin 12 behaviour and the
total rate are, within present errors, OK. Now pay attention: if
the production mechanism is only for the "wanted" lepton T and its
own neutrino, the final states would be:
e + ~+ + neutrinos
e + e+ + neutrinos
~ + (hadrons)+ + neutrinos
e + (hadrons)+ + neutrinos
The experimental results can be expressed as follows:
ee
~p
1.0 0.3 ,
ee
e~
~~
e~
0.5 0.3 ,
e + tracks
~ + tracks
1.0 0.3 .
The comparison of the experimental data (5.2) with three possibili-

2/
ties, Le. (T, 'J T ), (E, M), q 3, lS shown in Table 5.2. So far,
all possibilities remain open if we limit our investigation only
to (e+e-) physics. In fact "gauge leptons" can be excluded from
neutrino experiments up to masses of a few GeV.
Table 5.2
Experimental data compared with various theoretical possibilities
ee ee ww e + tracks
wii
-ew =
ew W + tracks
Experimental data 1 0.3 0.5 0.3 l.0 0.3
p
Heavy lepton with
0
.~
its own neutrino; 1 0.5 1
.j..J
ctt
as In SU(2) x U(l)
.j..J
u
Q)
p..
Two "gauge leptons"
~
Q)
E, H, quasi-degenerate 1 0.75 1
in masses
H
0
Q)
q
2/3 -+q%
.c + V + e 1 0.5 1
E-! e
6. Do We Understand the Inclusive Lepton Production

In Hadronic Interactions?
The most interesting effect reported 4 ) is the fact that the

average transverse momentum of (W+W-) pairs stops increasing with
H(W+W-) at '" 3 GeV. The trend up to H(W+W-) '" 3 GeV indicated a
steady increase. This is now changed. For M(W+W-) ~ 4 GeV the
transverse momentum of the (W+W-) pair reaches a plateau value of
12 GeV (see
Fig. 6.1). Notice that the naive parton model could not account
for those high values of PT. Notice also that the "parton" model
predicts scaling. But scaling is-expected to be violated in QCD
and it is found to be violated in lepton-hadron deep inelastic
scattering. So it could be that the observation of high PT in
large-mass lepton pairs is connected with scale violation observed
in lepton scattering. It would be of great value, eventually using
QCD, if a theoretical link between these two effects could be esta-
blished: this is needed because deep inelastic lepton scattering
is space-like, while (W+W-) pair production is time-like.
1196 A. ZICHICHI
2.0
1.6
A 1.2
I- 400
a
v 0.8
CPR
1
eAnderson et 01.
ee
CF -Hom et 01.
0.4
0
0 4 8 12
Q (GeV)
Fig. 6.1 The average transverse momentum (PT) of ~ pairs

versus the pair mass
Another important point in large-mass lepton pair production

is to reduce the uncertainties in the comparison between theory
and experiment to better than a factor of 3. This would allow us
to establish, in this field of physics, whether the colour degree
of freedom is present. In fact if colour exists, only equally
coloured quark-antiquark pairs can annihilate. This reduces, by a
factor of 3, the standard Drell-Yan process, because only one
colour out of three has the chance to meet the right quark colour
to annihilate into a colourless photon.
Figure 6.2 shows a summary of the CERN-ISR and Ferrnilab data

on inclusive prompt lepton pair production in the "continuum". A
fit to the continuum dimuon spectrum in the mass range
6 < M < 12 GeV gives
WW
A e-bM
with
A (1.89 0.23) x 10- 33 cm 2 /GeV/nucleus
b (0.98 0.02) GeV- 1
The datal;) show the same scale behaviour as (IN) inclusive produc-
tion (Fig. 3.1). A comparison with the data of Section 3 suggests
that (IN) and "continuum" corne from similar processes.
Concerning "single" prompt lepton production, its status ~s as

follows:
i) The energy dependence, i.e. the threshold effect at IS ~ 6 GeV,

is still contradictory.
ii) For PT > 1 GeV/c the bulk of single-lepton production (> 75%)
is due to pairs. However, the production of known vector
mesons (p, w, ~) is not enough. An important contribution
from the continuum is needed.
1198 A. ZICHICHI
-32
10
~
.8CSY
N
CP-I
~
l?
-33 . .... CFS
N
I
E
10 .
0
u ~I't
.'+'.~
"
>-
>-
-34
.~
"U
~
10
M~
"U T
......
N b
II
"U
'"~
1035
-36
10
0.1 0,2 0,3

M/..jS
004 0,5 0,6
Fig. 6.2 A summary of CERN and Fermilab data on inclusive

lepton pair production in the "continuum"
iii) The only firm evidence for prompt single leptons comes from a
recent Fermilab experiment where direct single-muon production
has been observed with PT < 1 GeV/c and at a rate comparable
to that of ~ pairs 19 ) . The origin of these prompt single muons
~s believed to be "charm" production.
7. Do We Understand High-Energy Neutrino Reactions?
A senes of accurate data from CERN (WAI group) shows that
i) there ~s no high-y anomaly;
ii) there is no threshold effect in (y);
iii) the ratio o(v)/o(v) does not rise with Ev.
The comparison of these data with previous Fermilab results is pre-

sented in Figs. 7.1 and 7.2.
The present status of high-energy neutrino physics is therefore

the following:
Charged currents 20 )
i) (y) is independent of energy;
ii) o(v)/o(v) is independent of energy;
iii) (y) '"

V O.S} as predicted from the naive quark model.
iv) (y)v '" 0.3
Neutral currents 21 )
i) The cross-section for neutral currents is proportional to

energy, 0NC '" Ev to better than 10%.
ii) The y-dependence ~s measured to be of the form
do
--
dy
= a + bel - y )
2
for v,
do 'iJ.,
-- = a(l for
dy
with a = 0.29 0.01; b = 0.06 0.01.
~
o
o
T T
1/
051- 8=0,8
--=~~=f=--
~.+ --+-
I I
>- 0.41-
.HPWF1
1.01- .. " " 2 -
CERN WA1
CERN -
0.31-
I I I
0,51- 1/
-
HFWF
WA 1
~ tit-~
~ 05~ k ~ I -+--+--+-+-+
f
~
~O+ !+~~=HI I~
0.3f- +-=F -
I I I 01 I I I 1
o 50 100 150 o 50 100 150 200
E-1/ Ev. GeV
Fig. 7.1 y-dependence versus neutrino or anti- Fig. 7.2 Comparison of the ratio between the
cross-sections of antineutrinos and neutrinos 1>
neutrino energy as measured at CERN (WAI Group) N
and Fermilab (HPWF) versus energy as measured at CERN (WAI Group) ()
and Fermilab (HPWF) :c
(')
:c
iii) C-symmetry is verified to better than 10%.
iv) The Salam-Weinberg weak angle is measured to be
sin 2 8 = 0.25 0.03.

w
Figure 7.3 shows a set of experimental results on 8 , in-

w
eluding the last one from CERN-~vAl.
3. Do We Understand Total, Elastic,

and Differential Cross-Sections?
Two new and interesting results show the existence of long-

range correlations:
a) one is relative to PT conservation, which is found by the

CERN-Bologna group22) to extend over a very long range, as far
as 5 units of rapidity, as can be seen from Fig. 8.1;
b) the other refers to another important quantity, the "multi-

plicity". The Aachen-CERN-Heidelberg-Munich Collaboration 23 )
has found that the multiplicity observed in the backward (n B)
and forward (n F ) hemispheres are clearly correlated, as shown
in Fig. 3.2.
Both results have been obtained at the CERN ISR.
9. Do We Understand the High-Energy Phenomena

Observed in Nuclei?
New results have been obtained 24 ) by Cronin et al. at Fermilab

. numb
on t h e atom~c Ad
er, - epen d ence, 0 f ( n-K-p-
+ + +).~nc 1us~ve
. processes.
a
The exponent a of A goes from a ~ 0.9 at PT = 1 GeV/c to a ~ 1.2
at PT ~ 4 GeV/c. Remember that low-mass lepton pairs (M~ ~ 1 GeV)
are produced in nuclei with a ~ 13. For M~~ ~ 3 GeV, a ~ 1. This
phenomenon is explained in terms of "wee" partons interacting
strongly and fast partons interacting weakly. A "wee" pair with
MI~ ~ 1 GeV annihilates into a virtual photon on the surface of the
nucleus. Otherwise a member of the pair would interact strongly
1202 A. ZICHICHI
Fig. 7.3 Comparison of Rv and R~ with the

Salam-Weinberg model. The dotted line assumes
no antiquarks; the solid line is for
q/q = 0.1 and corrected for the non-
isoscalarity of the target.
en
Unlike pairs
- (-+ +)
..
~
-;
C
en
10P'+ o"T1
e-t:.Y/3 -;
:::t:
.. m
en
C
ttl
5 Z
r'l'" "".'~+_+__+_-+__+_-+_
o C
(')
r
m
>
::0
-51-
I .- - ~
3 4 5 6 t:.y (n F ) :::t:
-<
en
11
like pairs (~!) Vs =63GeV

9
slope; .31
10~
.. + e -t:.Y/3
7
,--.-"t ,- -+--+-1 -+_ .LL _ n8 I n

~':I'. F
--;: ....~
5
(n.>
I 3'
3 4 5 6 5 10 15 20
t:.y
Fig. 8.1 Asymmetry parameter as a function of Fig. 8.2 Showing the correlation between the
6y for a) unlike pairs, b) like pairs. Notice average charged multiplicities in the forward
that the asymmetry extends to very high values and backward hemispheres in (pp) collisions
of 6y; and this is the evidence for "long- at IS = 63 GeV
range" effects. o
'"
Co)
1204 A. ZICHICHI
and lose its chance to annihilate. Pairs with M!~ 1 GeV can be
made up with fast partons. These pass almost freely through the
nucleus and can annihilate anywhere in it; at the surface or deep
inside. The A-dependence of the lepton pair cross-section should
vary with A~3 for M ~ 1 GeV and become proportional to A for
M 1 GeV. The interpolation between these two limits depends on
the details of quark interactions. Notice that for large Pr ha-
. processes, such as the ab
dron1c '
ove-ment10ned ( n-K-p-
+ + +).1nc 1US1ve
.
production, the A-dependence could have a power larger than 1 be-

cause of mUltiple scattering. This is not the case for lepton pair
production, whose A-dependence should saturate at the linear level.
The data follow this trend; the model outlined above is the one I
like most and is due to Farrar 25 ).
10. Do We Understand the Inclusive Electromagnetic

Coupling of the Hadrons?
Nothing new.
11. Should We Believe in Quarks?
The news here is the world analysis by the CERN-Bologna

group26 ) of all quark searches performed so far. The results are
shown in Fig. 11.1. All experiments are with Pr < 1 GeV/c. So,
if the production process for quarks needed, for example,
Pr 2 GeV/c, all these experiments would be meaningless.
Only
27
very recently Cronin et al. ) have performed an experiment to look
for inclusive quark production with Pr ~ 2 GeV/c (for q = 'l3) and
Pr ~ 4 GeV/c (for q ~3) with null results. As emphasized
earlier, the "strong" channel for quark production is affected by
the doubt that when a high Pr value is observed, this does not
necessarily correspond to a single transfer of Pr to the wanted
quark. It is in fact very likely that the Pr transfer process is
multiple. This means that the observed Pr is the sum of many
smaller Pr transfer processes. If there is a threshold effect 1n
quark production, it is important that the "observed" Pr really
15
10
0:-
a
0
a.le
--
z 3
as
~
I /)
C
GI
E
.~
a.le
8.5
~
3
w
o .2 .4 .6 .8 1.0
(GeV/c)
Fig. 11.1 Showing the values of PT for all quark
experiments. Note that 16 measurements have been
made at PT = 0.1 GeV/c and none above PT = 1 GeV/c.
The latest results of Cronin et ale are not in-
cluded. See text.
1206 A. ZICHICHI
corresponds to a unique transfer process. This is so when "hard"

probes are used, as for example those provided by ''weak'' and
"electromagnetic" couplings. Here the experimental observation of
a PT clearly corresponds to a unique transfer process. The con-
clusion is that the quark search should be continued using all
interactions (not only the strong ones as mainly done so far) and
investigating all possible configurations in the final states. For
example, an isolated quark search is unlikely to produce positive
results because, in a process which breaks a proton into pieces,
it is physically very sound to expect a large number of "hadrons"
in the final state, these being the "materialization" of the ha-
dronic cloud which surrounds the quark. Nobody knows how to co~
pute the transition from the virtual cloud state to the real one.
If this required, say, 20 GeV, the quarks could be very light;
still, in order to break a proton, a threshold energy of 20 GeV is
necessary.
The greatest enemy of quarks is geometrodynamics 28 ) , whose

basic point is that quarks are "coordinates" of hadrons and are
totally deprived of physical meaning: geometry generates dynamics
and not vice versa. This theory is very interesting for its very
rich predictive power: mass spectra, Regge trajectories, etc.
The most impressive prediction is, to me, the absence of the

P
20-plet for baryons, irrespective of the L state; and the absence
of the 70-plet corresponding to LP = 0+. As we know, these are the
weakest multiplets in the large variety of states discovered so far.
The theory is of the greatest interest because it can be disproved
in very many ways, thanks to the impressively numerous predictions.
For example, one single state of the (70, 0+) would be enough to
kill it.
12. Should We Believe in

Another Fundamental Interaction?
No, if nature has at least six quarks and six leptons, as dis-
cussed in Section 4.
13. Do We Understand the Universality Features Observed in Nature?
No.
14. Should We Believe in Gauge Theories?
Personally I like gauge theories, since our world is made of

finite quantities and we cannot go on with infinities allover the
place.
From the theoretical standpoint, however, we have to say "Wait

and see", as shown by 't Hooft's lectures 29 ). QCD, he said, would
have been a perfect theory if it did not have an important short-
coming: up to now it is only formulated as a (divergent) pertur-
bation expansion in a coupling constant g, which, in general, is
not small. 't Hooft's lectures were devoted to a possible scheme
that may eventually convert this expansion into a convergent pro-
cedure: after defining the coupling constant precisely, he finds
in the massless theory the singularities for complex values of this
constant. Then the Borel re-summation procedure is introduced.
There the perturbation expansion is governed by different types of
singularities, which are explored for different theories. Some are
governed by instantons, others by the renormalization group. In
spite of all these efforts, a convergent formalism is missing.
15. Can Internal and Lorentz Group Symmetries

be Unified?
Nothing new. For developments see Section 17.

1208 A. ZICHICHI
16. Should We Believe in QFT?
Wightman 30 ) has answered Yes. He started by recalling what it

means to believe in a physical theory; the old quantum theory, the
Boltzmann equation, the classical Maxwell theory of electricity and
magnetism, and quantum mechanics were discussed as examples. The
following definition was offered: one believes in a mature physi-
cal theory if a) it is internally consistent, b) there is a signi-
ficant range of experiments which it describes accurately. For a
"theory that ~s still being worked out", Wightman suggested that
the question of belief is one of betting-odds: how likely is it
that when the theory becomes mature it will be believable? Infor-
mation of both a theoretical and an experimental nature is relevant
in assessing such odds.
The application of these definitions to quantum field theory

(QFT) began with a review of the early history of the subject. It
was pointed out that the founders of QED did not believe in it, and
that even after the modification of the foundation of the theory
provided by the renorrnalization procedure, scepticism persisted.
Nevertheless, the remarkably successful experimental verifications
of the QED of muons and electrons have made it clear that there
must be something right about the theory.
Axiomatic field theory and constructive field theory have pro-

gressed far enough so that it is clear, for a wide variety of super-
renormalizable models, that solutions exist with reasonable physi-
cal properties. The most important open problem at the moment is
to obtain analogous results for renorrnalizable theories.
To all theorists who work in less than four dimensions we

should, however, recall that there are no spaghetti in less than
four dimensions.
17. Beyond the Present Knowledge
As stated in the opening lecture, one of the problems here is

to see if quaternions or octonions can lead to a deeper understanding
of particle physics. Not only complex numbers which are the
basic ingredient of quantum mechanics -- but also quaternions seem
inadequate to describe the variety of phenomena observed in sub-
nuclear physics.
It seems that we need to jump into the octonionic Hilbert

space if we want to unify the description of lepton and hadron
symmetry properties. The development of these ideas is not very
well known, and I would like to recall the main steps from Gursey's
fascinating lectures 31 ).
Octonions were brought into physics in 1934 by Jordan,

von Neumann, and Wigner, when they constructed an unconventional
quantum mechanical space associated with a finite number of degrees
of freedom. The construction involved triplets of octonions
naturally associated with a group SU(3) x SU(3) embedded in an ex-
ceptional group (F 4 ). No physical interpretation could be found
for these degrees of freedom and the discovery was forgotten. In
the meantime, mathematicians extended their model to larger excep-
tional groups, for example (E 7 ), SU(6) x SU(3). It is thanks to
Gursey and collaborators that the similarities between these spaces
and the transformation properties of quarks and leptons were noted:
a) the colour group SU(3) is associated with octonions and

c
cannot be broken (see Section 4);
b) the other SU(3) or SU(6) groups can be interpreted as flavour;

hence leptons and quarks can be unified as octonion fields;
c) the fundamental representations of exceptional groups give the

number of quarks and leptons and their quantum numbers, thus
providing a theory of colour and flavour simultaneously;
1210 A. ZICHICHI
d) Non-coloured states lie in a conventional Hilbert space,

while coloured states have different new properties (such as
failure of the superposition principle) corresponding to basic
differences between leptons and quarks.
All this looks extremely interesting. It could be that the way

to unify leptons and quarks is by means of octonionic Hilbert space.
Conclusions
th
It was the 15 year of the School, and we wanted to focus the
program on all the problems we could think of32): some hot and new;
others old and quasi-forgotten, in spite of their great relevance
to our understanding of the laws of nature.
References
1) H.J. Lipkin, "Why is there charm strangeness colour and all

that?". This Volume.
R. Bizzarri, "Do mesons fill SU(3) nonets?". This Volume.
2) ABCLV Collaboration, Phys. Letters 78B, 342 (1978). Reported

at this School.
3) M. Chen, "New particle production ~n hadronic interactions".

This Volume.
4) A.J.S. Smith, "Review of lepton production in hadron-hadron

collisions". This Volume.
5) R.A. Salmeron, private communication.
6) H.F. Schopper, "The properties of charmonium and charm par-

ticles". This Volume.
7) B.H. Wiik, "Recent results from DASP". This Volume.
S) K. Wilson, Phys. Rev. 179, 1499 (1969).
9) S. Nilsson, "Narrow resonances in BB reactions". This Volume.
10) N. Cabibbo, "Quark and lepton mixing". This Volume.
11) S. Weinberg, Phys. Rev. Lett. 19, 1264 (1967).

A. Salam, in "Elementary particle theory", Nobel Symposium
(ed. N. Svartholm) (Almqvist and Wiksell, Stockholm, 1963),
p. 367.
12) P.W. Higgs, Phys. Lett. 12, 132 (1964); Phys. Rev. Lett. 11,
508 (1964); Phys. Rev. 145, 1156 (1966).
13) O.W. Greenberg, Phys. Rev. Lett. 13, 598 (1964).

W. Bardeen, H. Fritzsch and M. Gell-Hann, in "Scale and con-
formal symmetry in hadron physics" (ed. R. Gatto) (J. Wiley
and Sons, Inc., New York, 1973), p. 139.
S. Weinberg, Phys. Rev. Lett. 31, 494 (1973).
H. Fritzsch, H. Gell-Hann and H. Leutwyler, Phys. Lett. 47B,
365 (1973).
14) M.L. Perl, Results presented at a CERN seminar on Heavy Leptons,

1977
15) H.S. Schopper, private communication.

1212 A. ZICHICHI
16) M. Bernardini, D. Bollini, E. Fiorentino, F. Mainardi,

T. Massam, L. Monari, F. Palmonari and A. Zichichi, "A
proposal to search for leptonic quarks and heavy leptons
produced by ADONE", INFN/AE-67/3, 1967.
17) M. Bernardini, D. Bollini, P.L. Brunini, E. Fiorentino,

T. Massam, L. Monari, F. Palmonari, F. Rimondi and
A. Zichichi, Nuovo Cimento l7A, 383 (1973).
18) J.S. Bell and R. Jackiw, Nuovo Cimento 60, 47 (1967).

S.L. Adler, Phys. Rev. 177, 2426 (1969)-.-
19) B. Barish, private communication.
20) J. Steinberger, private communication.
21) H.P. Paar, "Heasurement of neutral current cross-sections and

their energy- and y-dependence". This Volume.
22) 11. Basile, G. Cara Romeo, L. Cifarelli, P. Giusti, T. Massam,

F. Palmonari, G. Valenti and A. Zichichi, Nuovo Cimento 39A,
441 (1977).
23) S. Uhlig, 1. Derado, R. Meinke and H. Preissner, "Observation

of charged particles correlations between the forward and
backward hemispheres in pp collisions at ISR energies",
HPI-PAE/Exp.El. 66 (1977).
24) D. Antreasyan, "Nuclear effect observed in high transverse

momentum hadron production". This Volume.
25) G. Farrar, Phys. Lett. 56B, 185 (1975).
26) H. Basile, G. Cara Romeo, L. Cifarelli, P. Giusti, T. Uassam,

F. Palmonari, G. Valenti and A. Zichichi, Lett. Nuovo
Cimento~, 529 (1977).
27) D. Antreasyan, G. Cocconi, J .H. Cronin, H.J. Frisch,

L. Kluberg, J. Mueller, K. Olive, P.A. Piroue, M.J. Shochet
and R.L. Sumner, Phys. Rev. Lett. 39, 513 (1977).
20) G. Preparata, "Quark-geometrodynamics: A new approach to

hadrons and their interactions". This Volume.
29) G. 't Hooft, "Can we make sense out of 'quantum chromodynamics"'.

This Volume.
30) A.S. Wightman, "Should we believe in quantum field theory".

This Volume.
31) F. Gursey, "Symmetries of quarks and leptons". This Volume.
32) A. Zichichi, Opening lecture: "The whys of subnuc1ear physics".

This Volume.
CLOSING CEREMONY
The Closing Ceremony took place on Tuesday, 9 August 1977.

The Director of the School presented the Prizes and Scholarships
as specified below.
PRIZES AND SCHOLARSHIPS
Pnize 6o~ B~~ S~uden~ - awarded to Ashton B. Carter - Oxford

University, Oxford, UK.
The following students received hono~y menton6 for their

contributions to the activity of the School:
W.B. ATWOOD - CERN, Geneva, Switzerland.

J.G. BRANSON - Princeton University, Princeton, NJ, USA.
M. DINE - Yale University, New Haven, CT, USA.
J.H. FIELD - CERN, Geneva, Switzerland.
R. FISCllliR - Universitat Karlsruhe, Karlsruhe, FRG.
P. JENNI - CERN, Geneva, Switzerland.
M. JUNG - CERN, Geneva, Switzerland.
J. ~lOBLOCK - CERN, Geneva, Switzerland.
L. MANDELLI - Universita di Milano, Italy.
L. ROLANDI - INFN, S. Piero a Grado (Pisa), Italy.
K. TOTH - Hungarian Academy of Science, Budapest, Hungary.
P.K. TOWNSEND - State University of New York, NY, USA.
W. WETZEL - Heidelberg Universitat, Heidelberg, FRG.
Fourteen Scholarships were open for competition among the

participants. They were awarded as follows:
P~ck M.S. Btack~ Scholarship - awarded to Willia~

CELMASTER, Harvard University, Cambridge, MA, USA.
Jam~ Chadw~ck Scholarship - awarded to Ian HINCHLIFFE,
Oxford University, Oxford, UK.
Amo~-de Shait Scholarship - awarded to Hans REITHLER,
Aachen University, Aachen, FRG.
1215
1216 LIST OF PRIZES AND SCHOLARSHIPS
Gunn~ Kallen Scholarship - awarded to Hans P. PAAR,

CERN, Geneva, Switzerland.
An~e Lag~ue Scholarship - awarded to Dikran ANTREASYAN,
University of Chicago, Chicago, IL, USA.
G~o Racah Scholarship - awarded to Sau Lan k~,
University of Chicago, Chicago, IL, USA.
Gio~gio Ghigo Scholarship - awarded to Robert V. BUDNY,
Rockefeller University, New York, NY, USA.
Ennico P~ico Scholarship - awarded to Gideon BERLAD,
Technion, Haifa, Israel.
Pet~ P~~w~k Scholarship - awarded to Emil MOTTOLA,
Columbia University, New York, NY, USA.
Gianni Q.~eni Scholarship - awarded to Thomas J. WEILER,
University of Liverpool, Liverpool, UK.
Antonio S~anghettini Scholarship - awarded to Marjan BACE,
Heidelberg University, Heidelberg, FRG.
Albvr.:to Tomaoini Scholarship - awarded to Anthony I. SANDA,
Rockefeller University, New York, NY, USA.
E;t;to~e Majo~a Scholarship - awarded to Marek JEZABEK,
Krakow University, Krakow, Poland.
Benjamin W. Lee Scholarship - awarded to Alberto DE LA TO~,
Wuppertal University, Wuppertal, FRG.
Prize for Be6~ ScJ..em6ic SeM~Y - awarded to Charling TAO,

Harvard University, Cambridge, MA, USA.
The following participants gave their collaboration in the

scientific secretarial work:
Dikran ANTREASYAN Enore GUADAGNINI Anthony I. SANDA

Marjan BACE Bernard JANCEWICZ Bruce M. SCHECHTER
Laurent BAULIEU Marek JEZABEK Franco STROCCHI
Henrik BOHR Jochen KRIPFGANZ Charling TAO
Robert V. BUDNY Mihai 1 MINTCHEV Paul K. TOWNSEND
Ashton B. CARTER Emi 1 MOTTOLA Tsuneo UEMATSU
William CELMASTER Stephen L. OLSEN Klaus WACKER
Alberto DE LA TORRE Hans P. PAAR Bruce M. WEEKS
Frank J. DE LUCCIA Giampiero PAFFUTI Thomas J. WEILER
Michael DINE John P. PRESKILL Sebastian N. WHITE
John H. FIELD Hans RE ITHLER Sau Lan WU
Richard FISCHER Luigi ROLANDI
LIST OF PARTICIPANTS
Dikran ANTREASYAN University of Chicago

The Enrico Fermi Institute
5360 Ellis Avenue
CHICAGO, II 60637, USA
William B. ATWOOD CERN

EP Division
1211 GENEVA 23, Switzerland
Marjan BACE Institut fur Theoretische Physik

der Universitat Heidelberg
Philosophenweg 16
6900 HEIDELBERG, FRG
Rinaldo BALDINI FERRIOLI CELIO Laboratori Nazionali di Frascati

Casella Postale 13
00044 FRASCATI (Roma), Italy
Laurent BAULIEU CNRS Universite de Paris-Sud

LPTENS
24 rue Lhomond
75231 PARIS CEDEX OS, France
BERGSTROM Lars The Royal Institute of Technology

Department of Theoretical Physics
10044 STOCKHOLH 70, Sweden
Gideon BERLAD Technion

Israel Institute of Technology
Department of Physics
HAIFA 3200, Israel
Ikaros BIGI Max-Planck-Institut fur Physik

und Astrophysik
Institut fur Physik
Fahringer Ring 6
8000 MUNCHEN 40, FRG
Romano BIZZARRI Istituto di Fisica dell'Universita

Piazzale delle Scienze 5
00185 ROMA, Italy.
Henrik BOHR K6benhavn Universitet

Niels Bohr Institutet
Blegdamsvej 17
2100 K0BENHAVN 0, Denmark
1217
1218 LIST OF PARTICIPANTS
James G. BRANSON Princeton University

P.O. Box 708
PRINCETON, NJ 08540, USA
Robert V. BUDNY The Rockefeller University

1230 York Avenue
NEW YORK, NY 10021, USA
Burkhard BUNK Max-Planck Institut fur Physik

und Astrophysik
Institut fur Physik
Fahringer Ring 6
Nicola CABIBBO Istituto di Fisica dell'Universita

00185 ROMA, Italy
William CAMERON Rutherford Laboratory,

HEP Division
CHILTOND, DIDCOT, Berks., UK
Ashton B. CARTER University of Oxford

12 Parks Road
OXFORD OXl 3PQ, UK
William CELMASTER Harvard IUni vers i ty

Lyman Laboratory of Physics
CAMBRIDGE, MA 02138, USA
Min CHEN Massachusetts Institute of

Technology
CAMBRDIGE, ~~ 02139, USA
Sidney COLEMAN Harvard University

Lyman Laboratory of Physics
CAMBRIDGE, l~ 02138, USA
Clive COMBER Rutherford Laboratory

HEP Division
CHILTON, DIDCOT, Berks., UK
LIST OF PARTICIPANTS 1219
Nicola CUFARO PETRONI Istituto Nazionale di Fisica

Nucleare
Sezione di Bari
Via Amendola 173
70126 BARI, Italy
Alberto DE LA TORRE Gesamthochschule Wuppertal

Fachbereich 8 - Physik
5600 WUPPERTAL 1, FRG
Carlo DEL PAPA CERN

EP Division
Frank J. DE LUCCIA Harvard University

Jefferson Physical Laboratory
Bibhuti B. DEO Max-Planck-Institut fur Physik

und Astrophysik
Institut fur Physik
Fahringer Ring 6
Martine DETREMl1ERIE CEA-CEN Saclay

DPh-SECB
Boite Postale No. 2
91190 GIF-SUR-YVETTE, France
Michael DINE Yale University

Physics Department
217 Prospect Street,
NEW HAVEN, CT 06520, USA
Carlo DIONISI CERN

EP Division
Nils J. ELVERHAUG Universitet I Bergen

Allegt. 55
5014 BERGEN-U, Norway
Denis FAVART Universite de Louvain

Institut de Physique Corpusculaire
Chemin du Cyclotron 2
1348 LOUVAIN-LA-NEUVE, Belgium
Ariovaldo FERRAZ DE CAMARGO Universite Paris VII

LPTHE
Tour 33 - ler Etage
2 Place Jussieu
75221 PARIS CEDEX 05, France
Maria Itala FERRERO Istituto di Fisica dell'Universita

Corso Massimo d'Azeglio 46
10125 TORINO, Italy
Fernando FERRON I Istituto di Fisica dell'Universita

00185 ROMA, Italy
John H. FIELD CERN

EP Division
Richard FISCHER Universitat Karlsruhe

Institut fur Theoretische Physik
Kaiserstrasse 12
7500 KARLSRUHE 1, FRG
Henrik FLYVBJERG K~benhavn Universitet

Niels Bohr Institutet
Blegdamsvej 17
2100 K0BENHAVN 0, Denmark
Edi GAVA Istituto di Fisica Teorica

dell'Universita
Strada Costiera 11
34100 TRIESTE, Italy
Jean-Lucien GLOOR Universite de Lausanne

Institut de Physique Nucleaire
BSP Dorigny
1015 LAUSANNE, Switzerland
Enore GUADAGNINI Scuola Normale Superiore

piazza dei Cavalieri
56100 PISA, Italy
Feza GDRSEY Yale University

Physics Department
217 Prospect Street
NEWHAVEN, CT 65020, USA
Franz HERZOG Institut fur Physik der

Universitat Basel
Theoretische Physik
Klingelberstrasse 82
4056 BASEL, Switzerland
Ian HINCHLIFFE University of Oxford

12 Parks Road
OXFORD OXl 3PQ, UK
Viktor HUNGERBUHLER CERN

EP Division
Franco IACHELLO University of Groningen

Kernfysisch Versneller Instituut
Paddepoel
GRONINGEN, The Netherlands
Bernard JANCEWICZ University of Wroclaw

Institut of Theoretical Physics
ul. Cybulskiego 36
50205 WROCLAW, Poland
Peter JENNI CERN

EP Division
Marek JEZABEK Jagellonian University

Institut of Physics
ul. Reymonta 4
30-059 KRAKOW 16, Poland
Kj ell JOHNSEN CERN

ISR Division
Manfred JUNG CERN

EP Division
Nicola JURISIC Universitad Central de Venezuela

Facultad de Ciencias
Departamento de Fisica
CARACAS, Venezuela
Michael KAROWSKI Freie Universitat Berlin

Institut fur Theorie der
Elementarteilchen (WE 4)
Arnimallee 3
1000 BERLIN 33, FRG
Richard D. KENWAY University of Oxford

12 Parks Road
OXFORD OXl 3PQ, UK
Ritva KINNUNEN University of Helsinki

Siltavuorenpenger 20,
00170 HELSINKI 17, Finland
Jurgen KNOBLOCH CERN
EP Division
Dieter J. Kocher CERN

EP Division
1211 GEKEVA 23, Switzerland
Paul M. KOOIJMAN CERN

EP Division
Jochen KRIPFGANZ CERN

TH Division
Marcel LEBRUN Universite de Louvain

Paul LEBRUN Universite de Louvain

Denis LINGLIN IUT (IN2P3)

Laboratoire de Physique des
Particules
B.P. 909,
75019 ANNECY-LE-VIEUX CEDEX, France
Harry J. LIPKIN The Weizmann Institute of Science

REHOVOT, Israel
Jean-Francois LUCIANI Ecole Normale Superieure

Laboratoire de Physique Theorique
24 rue Lhomond
75231 PARIS CEDE X OS, France
Luciano HANDELLI Istituto di Fisica dell'Universita

via Celoria 16
20133 MILANO, Italy
Nicolae MARINESCU Institut fur Theoretische Physik

der Universitat Heidelberg
Philosophenweg 16
6900 HELDELBERG, FRG
Andre MARTIN CERN

TH Division
Francesco MARZANO Istituto di Fisica dell'Universita

Piazzale delle Scienze, 5
00185 ROMA, Italy
Karl Hubert MESS CERN

EP Division
Hannu I. MIETTINEN FERMI LAB

Theoretical Physics Department
P.O. Box 500
BATAVIA, IL 60510, USA
Adriana MINGUZZI RANZI Istituto di Fisica dell 'Universita

Via Irnerio 46
40126 BOLOGNA, Italy
Mihail MINTCHEV Scuola Normale Superiore

Piazza dei Cavalieri
56100 PISA, Italy
Emil MOTTOLA Columbia University

P.O. Box 133
Giuseppe NARDULLI Istituto di Fisica dell 'Universita

Via Amendola 173
70126 BARI, Italy
Christian NEF Universite de Geneve

Departement de Physique
32, Boulevard d'Yvoy
Sigward NILSSON University of Stockholm

Vanadisvagen 9
113 46 STOCKHOLM, Sweden
Stephen L. OLSEN Rochester University

Department of Physics and Astronomy
ROCHESTER, NY 14627, USA
Hans P. PAAR CERN

EP Division
Giampiero PAFFUTI Scuo1a Norma1e Superiore

Piazza dei Cavalieri
56100 PISA, Italy
Tchavdar PALEV Bulgarian Academy of Sciences

Institute of Nuclear Research
Boulevard Lenin 72
SOFIA 1113, Bulgarian
Patrice PAYRE CERN

EP Division
Jean PEREZ-Y-JORBA Universite de Paris Sud

Laboratoire de l' Acce1erateur
Lineaire
Facu1te des Sciences
91405 ORSAY, France
David E.L. POTTINGER Niels Bohr Institute

B1egdamsvej 17
2100 COPENHAGEN 0, Denmark
Giuliano PREPARATA CERN

TH Division
John P. PRESKILL Harvard University

Jefferson Physical Laboratory
Subhash RAJPOOT Imperial College

Prince Consort Road
LONDON SW7 2AZ, UK
Seifa11ah RANDJBAR-DAEMI Imperial College of Science and

Technology
Blackett Laboratory
Prince Consort Road
LONDON SW7 2BZ, UK
R. Ronald RAU Brookhaven National Laboratory

Physics Department
UPTON, NY 11973, USA
Hans REITHLER III Physika1isches Institut

der Rheinisch-Westfa1ischen
Technischen Hochschu1e
5100 AACHEN, FRG
Jean Fran~ois RENARDY CEN - SACLAY

DPHPE-SECB
Boite Posta1e n. 2
91190 GIF-SUR-YVETTE, France
Peter RENTON University of Oxford

Nuclear Physics Laboratory
Keb1e Road
OXFORD OX1, 3RH, UK
Georg RESELE Eidgenossische Technische Hochschu1e

Seminar fur Theoretische Physik
Honggerberg
8049 ZURICH, Switzerland
Luigi ROLANDI Istituto Naziona1e di Fisica Nuc1eare

Sezione di Pisa
56010 S. PIERO A GRADO (Pisa), Italy
Carlo RUBBIA CERN

EP Division
Anthony I. SANDA The Rockefeller University

1230 York Avenue
Ryu SASAKI Max-Planck-Institut fur Physik

und Astrophysik
Institut fur Physik
Fehringer Ring 6
Bruce M. SCHECHTER Massachusetts Institute of

Technology
Heinz SCHNEIDER Max-Planck-Institut fur Physik

und Astrophysik
Institut fur Physik
Fahringer Ring 6
Hervig F. SCHOPPER DESY

Notkestieg 1
2000 HAMBURG 52, FRG
Jorge Aurelio SILVA College de France

Laboratoire de Physique Corpusculaire
11 Place Marcellin-Berthelot
75231 PARIS CEDEX 05, France
A.J. Stewart SMITH Princeton University

P.o. Box 708
Jarn STEHR Universitat Hamburg

II. Institut fur Theoretische
Physik
Luruper Chaussee 149
Franco STROCCHI Scuola Normale Superiore

piazza dei Cavalieri
56100 PISA, Italy
Char ling TAO Harvard University

Gerard 't HOOFT Rijksuniversiteit Utrecht

Institute voor Theoretische Fysica
Princetonplein 5
UTRECHT, The Netherlands
Lothar TIATOR Johannes Gutenberg University

Institut fur Kernphysik
Fachbereich 18
Postfach 3980
6500 MAINZ, FRG
Kalman TOTH Hungarian Academy of Sciences

Central Research Institute for
Physics
P.O. Box 49
1525 BUDAPEST 114, Hungary
Paul K. TOWNSEND State University of New York

Institute for Theoretical Physics
STONY BROOK, NY 11794, USA
Michael J. TUITE Dublin Institute for Advanced

Studies
School of Theoretical Physics
10 Burlington Road
DUBLIN 4, Ireland
Tsuneo UEMATSU Kyoto University

Research Institute for
Fundamental Physics
KYOTO 606, Japan
Galliano VALENT Laboratoire de Physique Theorique

et Hautes Energies
Tour 16
4 Place Jussieu
75005 PARIS, France
Leon VAN HOVE CERN

Gustav Adolf VOSS DESY

Notkestieg 1
Klaus WACKER DESY

Notkestieg 1
Bruce G. WEEKS University of Michigan

2035 Randall Laboratory
ANN ARBOR, MI 48109, USA
Thomas J. WEILER University of Liverpool

Department of Applied Mathematics
Theoretical Physics
P.O. Box 147
LIVERPOOL L69 3BX, UK
Werner WETZEL Institut fur Theoretische Physik

Philosophenweg 16
6900 HEIDELBERG, FRG
Sebastian N. WHITE ~he Rockefeller University

.Department of Physics
1230 York Avenue
Arthur S. WIGHTMAN Princeton University

P.O. Box 708
Bjern H. WIIK DESY

Notkestieg 1
Jurgen WILLRODT Gesamthochschule Siegen

Fachbereich Physik
Postfach 21 02 09
5900 SIEGEN 21, FRG
Gaston WILQUET Universite Libre de Bruxelles

Inter-University for High Energies
ULB-VUB
Boulevard du Triomphe, C.P. 230
1050 BRUXELLES, Belgium
Robert R. WILSON Fermi National Accelerator

Laboratory
P.O. Box 500
BATAVIA, IL 60510, USA
Georg WOLF Max-Planck-Institut

fur Physik und Astrophysik
Institut fur Physik
Fehringer Ring 6
Sau Lan WU CERN

EP Division
Russell G. YATES University of Durham

Department of Mathematics
Science Laboratories
South Road
DURHAM DHl 3LE, UK
Georg ZOUPANOS University of Glasgow

Department of Natural Philosophy
GLASGOW G12 800, UK
INDEX
Abelian Higgs model Asymptotic freedom, 428, 632,

in 1 + 1 dimensions, 851 1061
Adler-Bell-Jackiw anomaly, scaling violation and, 603
860, 934, 1192 Axial currents, coupling with
Altarelli-Parisi analysis of QeD, pseudoscalar state, 765
604, 605
Anti-charm S states, 401 Bag models, 75, 729, 788, 1102
Antihyperons geometrodynamical, 730, 787
momentum distribution, 36 geometry of, 730
Anti-instantons, 814 MIT-Budapest, 73, 730, 787
Antineutrinos 0+ mesons, 195
charged current interactions quarks in, 142
with nucleons, 621 Barrier penetration, 807, 820,
neutral current cross sections, 882
673 Baryons, 17, 65, 108, 145, 732,
right-handed, 686 748, 798, 802, 1060
total cross sections, 607-634 angular momenta and, 793
asymptotically free gauge charmed, 23, 203, 211, 291,
theories, 624, 632 328
dimuon events, 631, 632 decay, 35
intermediate vector boson, evidence for, 293
624 colour, 45
narrow-band beam, 608 exchanging vector and
new flavour production, 624, pseudoscalar mesons,
666 1093
ratio to neutrino cross formation, 7
section, 5 high energy interaction, 774
results for charged current mass splitting, 21
events, 626 model of, 537
selection of event sample, molecular forces binding, 151
618 momentum, 36
wrong-handed, 684 production, 64, 286, 577
Antiprotons, 291 quark exchange, 1100
Antiquarks, 24, 1110 quark-quark interaction in,
in nucleons, 656 58
interaction with quarks, quark substructure, 534, 1073,
51 1181
1231
1232 INDEX
Baryons (cont'd) Bismuth (cont'd)

resonances, 148 Weinberg-Salam model and,
spin splitting of mass, 22 672, 687
strange, 291 Boltzmann equation, 984, 985
wave function, 748, 750, 791, Borel resummation procedure,
800 945, 955, 975, 978,
Baryon-antibaryon resonances, 52 996, 1011
Baryon-antibaryon systems, 67 Borel singularities
Baryon-baryon potential, 1072 complex plane, 965, 981
Baryon cross sections, 6 universality of, 958
Baryonium, 64, 68, 151, 395, 534, Bosons, 20, 1019, 1060, 1157
791, 1184 definition, 142
cross sections, 567, 576 fermion coupling, 1087
decay, 569 fundamental, 1062
definition, 537 interacting, 1044
exchange in pp annihilation, properties of, 1052
556 weak, 1062
formation, 570 Bounces, 808, 822, 885
isospin, 576, 579 construction of, 886
low energy bound systems and,
540 C* algebra, 1007
H:2 and MO' 578 Cabibbo angle, 3
off mass shell annihilation, Cabibbo structure, 295
542, 544 Cabibbo universality, 681, 682,
production, 573 684
quantum numbers, 540, 558 CERN Omega spectrometer, 553
resonances, 32 CERN 400 GeV super proton
below threshold, 569 synchrotron, 635
width, 571, 574 detector, 639
SU(6) symmetry, 751 Charged currents, 1199
trajectories, 559, 574 Charged current cross-sections,
BB reactions 644
masses, 1940-2500, 544 energy dependence, 664
masses, 2500- 00 , 548 ratio to neutral current
narrow resonances in, 533-579 cross sections, 644,
cross sections, 567, 576 658, 661
off mass shell annihilation, theoretical framework for,
542, 544 621
BEBC Charged V+A currents
muon identifier, 616 Cabibbo universality, 681,
narrow-band beam, 608 682, 684
neutrino flux, 611 in left-right symmetric gauge
selection of event sample, 618 models, 671-690
total cross-sections of lepton polarization, 679
neutrinos and rates for S decay and ~ decay,
antineutrinos in, 678
607-634 spectrum in ~ decay,
Binstantons, 848 679
Bismuth Charge exchange sum rule (CHIDO,
parity violation in, 722 115
INDEX 1233
Charm, 2, 3, 5, 20, 407, 1075, Charmonium state (cont'd)

1173 'postdictions', 347
behaviour, 23 properties of, 203-355
masses of particles, 220 pseudoscalar, 252, 258
splittings, 23 HI transitions, 255, 259
Charm burning, 3 production of, 41
Charmed states spectrum, 747 transition rates, 255
Charmonium, 1176 quantum numbers, 233
Charmonium spectroscopy, 24, 270, splitting, 223, 350
427, 431 total cross section and
Charmonium spectrum inclusive yield, 272
higher states in, 420 unbound resonances, 276
Charmonium state, 3, 27, 36, weak interactions, 295
37, 143, 790 sextet predominance, 305
annihilation graphs, 216 Charm particles, 203
bound states, 1177 Charm-strange exotics, 79
chromo dynamics , 211 Charm strangeness, 11
coupled decay channels, 216 Chiral symmetry, 1065, 1066,
decay, 32, 87, 229 1074
ny/n'Y and fy, 249 in extended Hoeller-Rosenfeld
E1 transition rates, 266 model, 1099
electromagnetic, 230 validity, 1076
experimental rates, 243 Chiral ward identities, 866
hadronic, 229, 230, 233, 257 Colour, 11, 44-69, 210, 728, 791
J/lj; and ljJ' 231 couplings, 63
leptonic, 231, 235 deuterons, 48
HI transitions, 271 gluon model, 53
mixing of cc states, 245 ~I=~ rule and, 147
of vector particles, 231 in multiparticle states, 53
OZI forbidden hadronic, 236 in nucleons, 48
OZI transitions, 251 oscillations, 92
P states, 260 quark confinement and, 45, 46,
possible modes, 229 48
products, 219 reasons fo r, 44
quantum numbers, 264 Colour degree of freedom, 18
radiative, 40, 243, 248 Colour spin(magnetic) exotics,
total width, 231 69
experimental, 204 Colour symmetry, 18
group theoretical models, 217 interference effect, 102
~ITS splitting, 228 Compton scattering in quark-
low-lying exotics, 86 parton model, 105, 106
of S-levels, 222 Conserved currents, 930
JP = 1- cc mass, 221 Constituent Interchange Hodel,
masses, 220 531
mass spectra, 218 Cosmic rays, 459, 639, 988
measurement, 218 contamination by, 639, 640
models, 210 Cosmology, 883
Niveau scheme, 221 Coupling constant, definition
P, splitting, 225 of, 946
positive parity, 30 Covariant derivatives, 828
1234 INDEX
Crazy mass formula, 21, 23 Dynamical symmetries in nuclear

Crazy parton model, 128 physics, 1043-1057
Cross-sections, 5, 1201
Cross section measurements, 157 Einstein's Riemannian manifold
for space time, 1066
DASP, recent results from, 357 Elastic cross sections, 1201
Deck mechanism, 183, 185, 186, Elastic scattering, 5
199, 201 Electrons
Deep inelastic processes, detection of, 475, 476
101-108, 685 neutral current coupling to,
parton model, 588, 590 673
quark distribution functions Electron-electron collisions,
in, 777 204
DESY, 41, 772 Electron-positron collisions,
Determinants, computation of, 204
899, 901, 924 antinucleon production, 291
Deuterons, 69 3D(3.77)-state, 276
composition of, 48 exclusive decays of 4.03 Gev
energy level spectrum, 48 resonance, 281
Dibaryons, 1184 fire sausage production in,
bound states, 75 772, 802
Dilepton pairs, mean transverse inclusive electron events, 371
momentum, 468 absolute cross section, 375
Dimensional renormalization inclusive electron yields, 285
scheme, 946 inclusive K-production, 283
equation, 949 inclusive n production, 287
Dimuons, 631, 632 inclusive Po production, 291
mass spectrum, 454, 457, 527 inclusive particle yields, 283
mass, relation to quark energy, photon production, 209
440 polarization of beam, 406
physics, 456 spectrometers, 206
production, 529 strange particle production,
Dimuon events, 459, 518 291
Dimuon final state, dimuon total cross sections, 272, 360
resonance in, 452 experimental results, 273
Dipion production, 167 higher resonances, 278
Diquarks, 52, 66, 67, 1110 kaon production and, 360
interactions, 149 Electron-positron storage rings,
Dirac sp in, 9 204
DORIS, 204 Electron-proton collisions,
beam energies, 346 1176
experiments with, 209 Elementary particles, 396
layout of, 205 hypothetical, 25
recent results from, 357 new, 395-433, 1173
spectrometer in, 207 in hadronic interactions,
Double well, instantons and, 435-470
812, 901, 920 production at large PT, 775
Drell-Yan model of quarks, 443, Equal opportunity quark model,
465, 466, 523, 528 45
test of, 513 Eta, 859
INDEX 1235
Euclidean Fermi fields, 862 Foldy-Wouthuysen transformation,

Euclidean functional integrals, 1075, 1078, 1082, 1087,
808 1101, 1147
Euclidean space, 1023 Fourier transformations, 841
four-dimensional, 808 Four particle correlations, 61
time ordered products in, 977 Free quark mass, 47
tunnelling amplitude of, 939
Euclidean space-time, field Gauge field theory, 8, 702, 828,
theory in, 956 1013, 1068
Euclidean two-space, 831 Abelian Higgs model, 837, 851
External muon identifier, 616 belief in, 1207
boundary conditions, 839
Fermions, 20 evaluation of determinant, 848
boson coupling, 1087 four-dimensional, 857
interacting, vertex symmetry in finite box of three-volume
for, 1087 V, 839
Fermi theory of weak interactions, infrared embarrassment, 848
987 sectors, 842
Field theory solution, 846
of quarks, 1074 splitting, 1152, 1154
unified. See Unified field unified theories and, 1115
theory vacuum structure of, 827
Figure of merit, 29 winding number, 829, 834, 837,
Finite action is zero measure, 845, 907, 921
906 Gauge group models, ambidextrous,
Fire-sausages, 766 674
angular momentum, 803 Gauge transformations, 828, 830,
cascade decay, 770 847
clustering, 771 Gell-l1ann-Low expansion, 1029
coherence in, 797 Gell-Hann-Nishijima form~la, 47
decay, 773, 775, 796, 798 Gell-Hann-Okubo mass formula,
logarithmic multiplicities, 21, 177, 743, 931
770 Gell-Hann-Zweig quark model, 93
production at high energy, Glimm-Jaffe bound, 1010, 1012,
775 1018, 1020, 1023, 1025
production in e+e-annihilation, Glueballs, 155, 572
772, 802 Gluons, 39, 1169
production in pp collisions, anomalous coupling, 425, 228
774 coloured model, 53, 1062
scaling, 770 coupling, 229
shape, 795 emission of, 587
short-range correlations, 770 exchange, 1067
structure, 798 for J/~ particles, 412
tilting probability, 776 exchange potential, 431
wave functions, 794 hard, 229
Flavours, 210 exchange of, 238
Flavour antisymmetry principle, in quark interactions, 211,
73, 147 1072, 1079
Fock space, vector, 52
1138 Gluon loops, 109
1236 INDEX
Goldstone boson, 859, 861, 875, Hadron-hadron collisions (cont'd)

932, 1157 lepton production in, 471-532
flavours, 931 Hadron loops, 110
mass of, 880 Hadron mass spectrum, 2
Goldstone dipoles, 862, 931 Hadron pair final states, new
Goldstone phenomenon, 928 particles in, 446
Goldstone poles, 875 Hadron physics, underlying
Graded Lie algebra, 20 structure, 1060
Gravitinos, 1068 Hadron scattering, high energy,
Gravitons, 1068 pomeron and, 771
Gravity, 938, 1111, 1154 Hadron spectroscopy, 26
unification with interactions, Hadron spectrum, low lying, 1060
1155 Hadron total cross sections
Group theory, 1043 flavour dependence, 119
regularities in, 119
Hadrons, 731 Hadronic beams, 41
Abelian quantum numbers, 13 Hadronic cross sections, 272
absorption of current, 101 Hadronic currents, behaviour of,
binding, 71, 73 297
charmed, 373 Hadronic decay, 233
coloured quark description of, of D mesons, 309, 312
55 of J/lji/1J;', 251
decay, 155, 373 of nc and n'c' 257, 259
electromagnetic coupling, 1204 of P-states, 268
exotics, 69 of upsilon family, 333
flavors, 931 OZI - forbidden, 236
forces between, 55 Hadronic final states particle
inclusive electromagnetic search in, 446
coupling, 6 Hadronic interactions
interactions, lepton production charmed particle production,
in, 5, 1195 451
isospin, 1067 J particle production in, 436
multiquark, 1184 cross sections, 444
nature of, 431 factors, 439
neutrino coupling to, 801 with different incident
production, 12, 63, 209, 371 beams, 443
properties, 26 new particle production in,
quantum numbers, 13 435-470
quark bound to, 47 search for charm production
quark structure, 1063 in, 468
saturation of structure, 7 Hadronic molecules, 732
Hadron extension in space time, Hadronic neutral current, 296
727 Hadronic scattering, high-energy,
Hadron final states 766
particle search in, in ~h~ Hadronic symmetry
mode, 450 approximate SU(6) x SO(3)
Hadron-hadron collisions, 275 orbital, 1076
lepton pair production, 496 in Moeller-Rosenfeld model,
lepton production, 497 1099
pair mass spectrum, 497 in potential approximation, 1093
INDEX 1237
Hadronic symmetry (cont'd) Instantons (cont'd)

internal and space-time, 1066 SU() solution, 851
involving spin and internal unstable states, 822
symmetries, 1101 with centre at tl' 813
mixing spin and internal Interactions
quantum numbers, 1069 number of, 9
quark substructure and, 1066 strong, 1064
. vertex symmetry for interacting unification of, 1104, 1112,
fermions, 1087 1136, 1154, 1156, 1160,
Han-Nambu model, 88, 93, 99, 105, 1193
137, 138, 140, 146, 1073 E6 scheme, 1114
Higgs mechanism, 1061 Isospin, 1, 7, 12, 1171
Higgs phenomenon, 852, 928
High energy spectroscopy, 24-36 J/W and W' family, 308, 329,
signal and noise in, 28 347, 395, 398-407, 472,
background signatures, 34 483, 497, 503, 746
production of low noise age, 478
signal, 30 decay, 231, 240, 250, 350,
High mass resonances, 34 352, 355, 402, 406, 1166
Hilbert space, 15, 1209 hadronic, 224, 234, 243, 251
octonions and, 1122 radiative, 243
properties of, 1063 width, 418
symmetry algebra description e.m. widths, 864
in, 16 G. parity, 399, 402, 403
Hypernuclei, 76 isospin, 399, 402
Hyperons, 22 nature of, 412
momentum, 36 production, 282, 463, 473, 511,
Hyperon nucleon cross sections, 1165
412 pseudoscalars, 414
quantum numbers, 399
Inclusive electron events, resonance parameters, 232
origin of, 371 scaling, 1174
Infinite space, 839 SU(3) classification, 399
Infinity, boundary conditions, transition rates, 415
922 theoretical discussion of, 411
Instantons, 9, 155, 805-941, 974, width, 518
1068 J particle, 27, 435
action of, 814, 843 detection of, 29
configurations, 917, 919 energy dependence of production
double well and, 812, 901, 920 cross section, 436
generalities, 843 production of, 27, 28, 30,
in particle mechanics, 808 464, 466, 468
interactions, 918 cross sections, 444
particulars, 845 factors, 439
periodic potentials, 820 on copper, 444
restoring symmetry, 925, 928 with different incident
singularities, 979, 981 beams, 443
singularities in F(z) due to, with different targets, 467
960 Jordan algebras, 1122, 1130,
size of, 814 1137, 1161
1238 INDEX
Jordan formulation, 1129 Lepton pair production (~ont'd)

Junctions, 576 dependence on beam energy, 515
dependence on beam particle
Kaons type, 513
charged, inclusive spectra of, dependence of target atomic
360 number, 510
masses, 20 dynamical features, 503
yields, 284 evidence for scaling, 515
resonant and non-resonant
Lattice gauge theory, 1078 pairs, 515
Leptons, 4, 1190 transverse momenta, 530
change into quarks, 1108 Lepton-quark symmetries, 1104,
decay, 272 1152
deep inelastic scattering, 18 Leptonic decay, 2, 231, 235
hadronically produced, 496 D mesons, 298
heavy, 4, 275, 364, 391, 1190 F mesons, 298
production, 374 K mesons, 610
in hadron-hadron collisions, leptons, 483
5, 471-532, 1195 upsilon, 378, 381
leptonic decay, 483 modes, 451
mass, 1155 Leptoquarks, 1110
momentum spectrum, 380 mass of, 1111
of large transverse momenta, Lie algebra, 827, 837, 1055,
473, 483, 505, 510, 518 1114, 1162
of transverse momenta, 478 graded, 1053
pair production, 607 Lipkin's crazy Parton model, 128
polarization of, 487, 679 Lorentz group symmetries, 9
production, 527 unification, 1207
role of, 4 Lumps, 813
search for, 472
sequential, 333 Massive quark model, 741
single prompt, 1197 Massive Thirring model, 1031
conclusions, 495 on-shell solution of, 1027
origin of, 483 l1ass splittings, 20
production of, 488 Matter
symmetries of, 1059-1164 quark structure of, 534
transverse momenta, 524, 530, Melosh transformation, 1075,
531 1086, 1101
unification with quarks, 1104 Me rolJ.s, 879
E6 scheme, 1114 Uesobaryon, 1184
symmetry breaking, 1119 Mesons, 17, 69, 203, 732, 1060,
yield of pairs, 488, 489 1178
Lepton mixing, 691-726 See also Quark-antiguarks
CP and T violation, 712 1935, 542
economy model, 705 0(980), 162
effects, 701 axial vector (Q), 110
muon electron transitions, 708 B (1235), 177
v-e transitions, 701 C-, 189
Lepton pair production, 532 charm and strangeness, 364
alignment in, 503 charmed, 3, 294, 327, 438
INDEX 1239
Mesons (con t' d) Hesons (cont'd)

decay, 1176 F+, 32
mass, 327, 1179 decay, 409
mass differences, 430 F*,
semileptonic decay, 374 evidence for, 43
semileptonic branching ratio, F, 286, 308, 392, 427
378 Cabibbo allowed non-1eptonic
charmed-anticharmed, 408 decays, 302
charmed-strange four quark, 81 decay, 283, 287, 306, 323, 348,
colour and quark charge, 90 664
confinement, 733 discovery of, 323
continuity, 733 evidence for, 43, 364, 411
coupling, 756, 760, 799 1eptonic decay, 298
cross sections, 6 production, 365
D+ semi1eptonic decay, 299, 319,
decays, 310, 311, 409, 410 322
mass, 430 splitting, 224, 228
D* filling SU(3) nonets, 159-201
decay, 312, 314 flavour singlet, decay, 97
mass, 318 forming four quark system, 69
production, 309, 314 geometrodynamical picture of,
D, 276, 282, "308, 4'J.7, 438 733
branching fractions, 311 interactions, 752
branching ratio, 286 K, 321
Cabibbo allowed non-leptonic decay, 610, 632
decays, 302 leptonic decay, 610
Cabibbo forbidden non-leptonic production, 283
decays, 304 mass, 328, 741, 744
decay, 283, 287, 306, 308, mass splitting, 21
348, 525, 664 maximum freedom, 734
hadronic decay, 309, 312 missing, 858
lepton decay of, 298 mixing, 108
mass, 316 mixing angles, 114
nonleptonic decay, 301 molecular forces binding, 151
production, 309, 314 non-leptonic decays, 305
production above 4 GeV, 312 non-trivial decay amplitude
semileptonic decay, 299, 319 relations, 1102
spin, 35 0+,
splitting, 224, 228 in MIT bag model, 195
DO 0++, as four quark states, 77
decays, 310 P, 435
mass, 430 parity violation, 410
mass spectrum, 355 pseudoscalar, 2, 108, 246,
decay, 229, 236, 327, 789 252, 258, 294, 414, 736,
muon radial dependence, 612 738, 741, 1092, 1093
dual system, 65 problems, 114
E(1420), 191 production, 153
exchanges, 117, 540 quantum numbers, 410
exotic, 62 transition rates, 255
F, 32 two-photon decays, 94
wave functions, 753
1240 INDEX
Mesons (cont'd) Muon number violation, 711

quark-antiquark interaction Uuon pairs, production, 513
in, 58 Huon towers, 207
quark coupling, 1092
quark substructure, 534, 1181 Neutral currents, 607, 1199
resonances, 533 charm-changing, 657
semi1eptonic decays, 308 coupling to electrons,
spectrum, 732, 736 673
structure, 7 in neutrino reactions,
tensor, 723
two-photon decay, Neutral current cross sections,
94 635-669
vector, 737, 738, 741, 758, charged current background,
799, 1071, 1092, 1093 643
wave functions, 753 charge symmetry invariance,
wave function, 735, 736, 752, 653
753, 757 charm changing, 657, 668
weak interactions, 295, 327 compared with charged current
Yukawa theory of, 987, 988 cross sections, 644
Meson-baryon interaction, 57 data reduction, 639
Meson collisions, 576 detector, 639
tlesonium, 1184 direction of hadron shower,
Minkowski space, single scalar 666
field in four dimensions, energy and y-dependence of,
805 648
Moeller-Rosenfeld model, chira1 event length cut, 641
symmetry in, 1099 Ev dependence of Nc/cc, 651
Ho1ecu1ar charmonium, 420 fiducial cut, 641
1101ecu1ar forces, binding by, 151 for neutrinos and
11u1ti-instanton, 918 antineutrinos, 673
Hu1tipartic1e states, 53, 732 hadron energy, 665
l1u1tip1ets, decay, 1076 Ke3 background, 645
l1uons neutral current sea and
charge asymmetry, 487 Heinberg-Sa1am model,
definition, 379 656
detection of, 207, 457, 475 ratio to charged current cross
distribution of intensity of, sections, 644, 658, 661
612 results, 646
from meson decay, 632 Rv and Rv without cut in
identification, 616 hadron energy, 653
number violation effects, 701 significance of right-handed
production, 371 component, 654
A-dependence, 529 total energy, 637
differential cross-section, wide-band beam background,
489 645
radial dependence from meson y-distribution, 651
decay, 612 Neutral current sea,
secondary production, 489, 492 Heinberg-Sa1am model
wrong-sign, 657, 668 and, 656, 658, 663, 667
yields, 379 Neutrinos
Muon electron transitions, 708 anomalies, 716
INDEX 1241
Neutrinos (cont'd) Nucleons (cont'd)

charged current interaction movement, 1044
with nucleons, 621 parton distribution in, 581
event length distribution, spin and isospin variables,
641, 642 1043
neutral current cross Yukawa theory of, 987
sections, 673 Nucleon-antinucleon interactions,
oscillations, 92, 694, 701 54, 70, 71
CP and T violation, 712 Nucleon-nucleon reactions,
production, 699 31, 54, 70, 1045
scattering, 801
total cross-sections, 607-634 Octonions, 1115, 1164, 1209
asymptotically free gauge Hilbert space and, 1122
theories, 624, 632 Octonionic quantum mechanics,
determination of flux, 611 internal symmetry and,
dimuon events, 631, 632 1122
intermediate vector boson, Off mass shell annihilation,
624 542, 544
narrow-band beam, 608 Off-shell quantities, 1040
new flavour production, On-shell solution of Eassive
624, 666 Thirring model, 1027
relation to antineutrino Osterwalder-Schrader theory,
cross sections, 5 977, 1023
results for charged current OZI rule violations, 354
events, 626
selection of event sample, P states, 414, 415, 429
618 decay, 260
wrong-handed, 684 El transitions, 266, 271
Neutrino beam at CERN 400 GeV hadronic branching ratios,
super proton synchrotron, 264
635 hadronic decay, 268
Neutrino events contamination by hadronic widths, 268
cosmic rays, 640 mixing with S states, 722
Neutrino flux, determination of, production of, 29
611 quantum numbers, 264
Neutrino reactions spin, 263
high energy, 1199 pp annihilations, 3
neutral currents in, 723 pp protonium states, 540
Neutron, 11 Partons, 1061
Nuclear physics, dynamical in constituents, 587
symmetries in, 1043-1057 in nucleon made of three
Nuclei, 6 constituent quarks,
high energy phenomena in, 1201 584
isospin of, 12 in QED, 590
mass number and charge, 12 scaling distribution, 602
Nucleons, 7, 48, 999 Parton distributions, 581-606
antiquarks in, 656 photon emission, 593
charged current neutrino and Q2 dependence, 587, 590, 603
antineutrino interactions, scope of model, 589
621 t dependence, 597
colour, 48 Parton models, 19, 137, 582, 984
interactions, 17 coherence in, 91
1242 INDEX
Parton models (cont'd) Pontryagin index. See tlinding

crazy, 128 number
dee~ inelastic phenomena, 588, Positronium, 229
590 cutoff in, 429
justification for, 583 force between atoms, 70
photon absorption in, ortho, 237
89, 91 Proton, 11
Perturbation theory, small constituent quarks in, 586
coupling approximation mass, 11
in, 806 stability, 10, 1110
Photons, 365, 1062, 1068 Proton-nucleus collisions
absorption, 105, 137 dimuon final states in, 452
absorption by coloured quark hadron pairs in, 446, 448
parton, 89, 91 Proton-proton cross section, 124
conversion into e+e-pairs, 476 Proton-proton scattering, 774
coupling, 229, 800 Pseudobaryon, 1184
decay of, 209 Pseudoscalar nonet, 742
production, 40, 209, 229 Pseudoscalar states coupling
quark absorption, 107 with axial currents, 765
of pseudoscalar and tensor
mesons, 94 Quantum chromodynamics, 211,
Physical sciences 727, 805, 858, 876,
belief in, 983, 1018 943-982, 1013
experimental validity of, 985 A1tarelli-Parisi analysis,
Pions, 7, 52, 999 604, 605
decay, 385 analytic structure for complex
exchange, 70, 71 g2, 952, 973, 975
dipion production and, 167 baby version, 868
masses, 20 Borel resummation, 945, 955,
momentum, 168 975, 978
production, 291 Borel signularities,
theory of nuclear forces, 1071 university of, 958
Pion-nucleon reactions, 31 coupling constant, 946
Pion-pion bound states, 420 definition, 943
Planck mass, 1111 difference with QED, 590,
PLUTO spectrometer, 207 975, 979
Poincare group, 1147, 1151 dimensional renormalization,
Pomeron, 153, 796 . 946, 949
definition, 121 dynamic structure of, 879
exchange, 794 free quark approximation,
flavour dependence in coupling, 1079, 1092
39 mass parameters in terms of
formation, 150 large momentum limits,
high energy hadronic 946
scattering and, 771 quark pair production in, 601
processes involving, 38 second Borel procedure, 967
slope, 6 singularities in F, 962
two-component formula, 122 singularities in F(z) due to
fits to higher energy, 124 instantons, 960
violation of OZI rule, 39 spin dependent forces in,
Pomeron diagram, 128 1067
INDEX 1243
Quantum chromodynamics (cont'd) Quarks (cont'd)

treatment, 862 charmed (cont'd)
validity of, 936 decay, 360
Quantum electrodynamics, 987, coherence, 88
1013 colour, 12, 62, 210, 728, 1136
divergence of, 993 charges, 90
Quantum field theory, 9, properties, 47
729, 983-1025 colour oscillations, 88
axiomatic, 1000, 1017, 1018, constituent
1208 in parton, 584
belief in, 983, 1208 p and n types, 586
comparison with experiment, coupling, 850
1013 decay modes, 82
consistent high-energy mass of, 427
behaviour, 1011 nature of, 435
constructive, 1003, 1017, replacing strange quarks, 22
1018, 1208 spin-dependent interactions,
early history of, 986 80
general theory, 1000 degenerate, 697
local, 1006 distribution of masses, 741
mathematical foundations, energy,
1009 relation to dimuon mass, 440
renormalization in, 989, 999, energy momentum, 1082
1000, 1006, 1009, 1022 evidence for, 7
super-selection sectors, 1008 exchange, 1100
Quantum mechanics, 985 field theory of, 1074
exceptional, flavours, 12, 210
invariance properties, 1130 flavour dependent mass term,
octonionic , 114
internal symmetry and, 1122 formation, 40
Quantum numbers, 12 four particle states,. 59, 69,
determination of, 233 72, 79, 537
historical, 1069 as molecule, 152
Quantum Sine-Gordon equation, charmed, 81
1027 flavour antisymmetry and,
Quantum theory, 838 147
Quarks, 13, 858, 1056 lines of force, 67
anomalous Pauli coupling, 215 0++ mesons as, 77
asymptotic freedom, 212, 353, with flavour antisymmetry,
1061 77
b, 721 free behaviour of, 735
belief in, 1204 Gell-Hann-Zweig model, 93
binding, 71, 73, 74 gluon interactions, 1079
bound to hadrons, 47 Han-Nambu model, 93, 99, 105,
change from leptons, 1108 137, 138, 140, 1073
charges, 88, 1081 colour in, 146
colour and, 90 heaviness of free, 51
measurement of, 88 heavy, 47, 421, 422, 747,
charmed, 18, 21, 24, 746, 1153 1078
annihilation corrections, interaction, 211
247 \Tith antiquarks, 51
cc bound systems, 191 with gluons, 1072
1244 INDEX
Quarks (cont'd) Quark geometrodynamics,

internal quantum numbers, 727-803, 1206
1104 at high energy, 766
Jordan algehra, 1134 baryons, 748, 798, 800, 802
left and right-handed, 1134 baryonium states, 791
Lorentz properties of L(r), current-particle matrix
215 elements, 760
making hadrons, 12 currents and wave function
mass, 52, 421 normalizations, 757
massive model, 741, 870 e.m. widths of J/~ family,
massless, 881, 1103 764
~eson coupling, 1092 freedom of, 731
mixing, 245 hidden and apparent-charmed
model for ~-, 49 states, 746
N-type, 694 meson interactions, 732, 752
number, 1081 principles, 731
pair production in QCD, 601 pseudoscalar nonet, 742
photon interactions, 107, 1105 simplicity of, 731
potential models, 213 three-meson coupling, 756
proton absorption, 137 weak interactions in, 801
scattering, 122 Quark-gluon interactions, 696
sequential, 333 Quark graphs for two-body
spectator, 153, 1093 nonleptonic decay, 301
standard model, 214 Quark mixing, 691-726, 1189
strange, CP violation, 699, 720
scattering, 122 possible CP violation, 694
substructure, 1065 production and decay of
symmetries of, 1059-1164 B particles, 698
t, 721 Quarkonium spectroscopy, 36
transitions, 7 flavour dependence, 37
transverse momenta, 778, 1204 production mechanisms, 38
unification with leptons, 1104 hadron production, 39
E6 scheme, 1114 new, 41
symmetry breaking, 1119 production mechanisms, 38
valence, 588 Quark parton model, 622, 729,
wave motion, 734 777, 794
weak interactions, 301, 1105 Compton scattering in, 105,
Quark-antiquarks, 36 106
binding, 71, 73 deep inelastic scattering,
Quark-antiquark annihilation, 466 101
Quark-antiquark pairs, behaviour SU;ll rules, 139
of, 46 Quark-photon interactions, 696
strange, 31 Quark-quark interaction, 52
Quark confinement, 8, 19, 45, 46, Quark structure of matter, 534
213, 729, 73-3, 735, 793, Quaternions, 1209
858, 879, 925, 933,
980, 1019 Reggeon diagram, 128
colour and, 45, 46, 48 Relativistic S matrix, 988,
Quark diagrams 1034, 1064
doubly disconnected, in 1 + 1 dimensions, 1027
240
singly disconnected, 237
INDEX 1245
Relativity, 731 Spectroscopy (cont'd)

Renormalization theory, 989, high energy. See High energy
1009, 1022, 1192 spectroscopy
essential nature of, 1006 Strangeness, 2, 20, 26, 121,
perturbation, 993, 999, 1000 1019, 1171
Renormalon, 981 oscillations, 92
Resonances, decay products, 219 splittings, 23
Rosner's theorem, 145 Strangeness exchange sum rule
Rotational invariance, 16 (SEX), 115
Strangeonium states, 31, 36, 427
S states, 722 Strong interactions, 727
anti-charm, 401 SU(2) symmetry, 827, 833, 837,
mixing with P states, 722 858, 1091
theory, 416 breaking, 932, 1185
Scaling, 1061, 1174 in QCD, 868, 878
violation, 605 limits, 859
Schrodinger equation, solution, SU(3) nonets
903 Al peaks, 177, 182, 183, 185,
Schwinger functions, 1012, 1022 189, 199, 200, 201
Schwinger model, 861, 928 1+ , 177
Sine-Gordon equation, 1027 I 0, 189
SLAC streamer chamber, 492, 550 I = ~, 174, 183
Solitions, 813, 1028 I = I, 177
calculation of form, 1040 KK system, 162, 164, 172, 196,
description, 1029 197
Soul, 570 masses, 177
Space, flatness of, 938 mesons filling, 159-201
Space-time 0+, 162
combined with internal I 0, 162
symmetry, 1160 I = ~, 174
Einstein's Riemannian manifold, I = I, 166
1066 0++, 196
Euclidean field theory in, members of, 160
956 SU(3) symmetry, 7, 17,49,108,
field theory, 955 234, 827, 851, 931,
four-dimensional, single/ 1060, 1072
scalar fields, 882 in QCD, 876
geometry of, 731 breaking, 118, 1185
two-dimensional, 929, 986 classifying observed spectrum
Space-time lattices, 944 of particle states,
Space-time regions. See Bag 159
models colour, 1061
SPEAR, 41, 204, 772 flavour, 1061
beam energies, 346 spectrum, 744, 745
magnetic detector in, 206 SU(4) symmetry, 18, 1070,
Spectrometers, 206 1078
double-arm, 208 breaking, 1097
Spectroscopy SU(5) symmetry, 1050, 1110
charmonium. See Charmon;_um extension of,
spectroscopy 1112
1246 INDEX
SU(6) symmetry, 17, 75, 586, U(l) symmetry, 831, 862

1046, 1053, 1060, 1063, breakdown of, 877, 880, 881
1065, 1066 changing vacuums, 872
baryon states, 751 chiral ward identities, 866
breaking, 76 Euclidean Fermi fields, 862
geometrical description, 1047 missing meson in, 858
limit, 1073 't Hooft's solution of, 858
regularities, 1076 Unified field theory, 1063,
supersymmetry and, 1104 1064, ll49, ll60
SU(3)c symmetry, 1189, 1209 Universality features in nature,
Summability methods, 995 8, 1207
Superalgebras, 1161 Universe
Supergravity, 1161, 1162 origin of, 898
Super-selection sectors, 1008 big bang theory, 883, 937
Supersymmetries, 1053, 1068, symmetry breaking in, 884
ll61, ll62 Upsilon, 472, 790, 1173
local, 1063 calculation of, 431
SU(6) symmetry and, 1104 decay, 336
unification and, 1155 excitation energies, 331
Symmetry experimental results, 329
chiral, 1065 hadronic decay, 333
combined with space-time, 1160 interpretation of, 331, 335
historical, 1069 leptonic decay modes, 378
internal and space-time, 1066 production and decay, 332, 503
involving spin, 1101 semihadronic decays, 381
lepton-quark, 1152 semileptonic decay, 392
octonionic quantum mechanics spectrum, 748
and, 1122 summary of information on,
unification, 1149, 1207 388
Symmetry algebras, 14, 15, 16 weak interactions, 382
Symmetry breaking, 3, 20, 109, width, 426
1l8, 877, 932, 933, 999,
1070, 1071, 1075, 1097, V+A currents, charged. See
ll50, 1185 Charged V~A currents
high temperature and, 878 Vacuums, 941
in early universe, 884 degenerate, 1150
patterns of, 1119 false, 882
restoration of, 925, 928 bounce, 885
U(l), 880, 881 collapse of, 937
decay of, 883, 897, 935
Thirring model, 1030 determinants and
massive, 1031 renormalization, 894
on-shell solution, 1027 fate of, 891
Total cross sections, 1201 thin-wall approximation,
Transverse momenta 888
final states, 780 rapidly expanding bubble wall,
particle production at, 893
775 true, 889
Triplet-singlet splitting, bubble of, 883, 892, 897
2 unstable, 882
INDEX 1247
Vacuum structure of gauge field Heinberg-Salam-l-lard-GU1 ll1Odel,

theories, 827 334
Vector currents, 758 \Ugner supermultiplet theory,
coupling with vector states, ]043, 1147
760 Wilson zero, 1010, 1022
Vector particles, 457 Ttlinding number, 829, 834, 837,
845, 860, 907, 921, 958
Wave functions spatial functions, t~B approximation, 807
59
Weak interations, 671, 801 x-onium pseudoscalars,
bounds on parameters, 680 detection of, 43
Fermi theory of, 987 x-onium states, 37
gauge theories, 691 X states, 227, 397, 400, 426
infinity problems in, 944 decay, 425
Weak neutral currents, 27 spin, 406
Ueinberg-Salam model, 647, 672,
719, 723, 884, 1061, Yukawa theory of mesons and
1106, 1191, 1201 nucleons, 987, 988
atomic parity violation, Yang-lUlls theory, 1061
724
neutral current sea and, 656, Zweig rule, 150, 352
658, 663, 667 Zweig-violating decays, 141

The Whys of Subnuclear Physics - Zichichi (Ed)

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

The Whys of Subnuclear Physics - Zichichi (Ed)

Загружено:

Авторское право:

Доступные форматы

The lIftys of

Series Editor: ANTONI NO ZICHICHI

I. 1963 STRONG, ELECTROMAGNETIC, AND WEAK INTERACTIONS

2. 1964 SYMMETRIES IN ELEMENTARY PARTICLE PHYSICS

3. 1965 RECENT DEVELOPMENTS IN PARTICLE SYMMETRIES

4. 1966 STRONG AND WEAK INTERACTIONS

5. 1967 HADRONS AND THEIR INTERACTIONS

6. 1968 THEORY AND PHENOMENOLOGY IN PARTICLE PHYSICS

8. 1970 ELEMENTARY PROCESSES AT HIGH ENERGY

9. 1971 PROPERTIES OF THE FUNDAMENTAL INTERACTIONS

10. 1972 HIGHLIGHTS IN PARTICLE PHYSICS

II. 1973 LAWS OF HADRONIC MATTER

12. 1974 LEPTON AND HADRON STRUCTURE

13. 1975 NEW PHENOMENA IN SUBNUCLEAR PHYSICS

14. 1976 UNDERSTANDING THE FUNDAMENTAL CONSTITUENTS

15. 1977 THE WHYS OF SUBNUCLEAR PHYSICS

PLENUM PRESS NEW YORK AND LONDON

Proceedings of the 1977 International School of Subnuclear Physics

1979 Plenum Press, New York

No part of this book may be reproduced, stored in a retrieval system, or transmitted,

From 23 July to 10 August 1977 a group of 125 physicists from

The countries represented at the School were: Belgium, Bulgaria,

The School was sponsored by the Italian Ministry of Public

The School was very exciting due to the impressive number of

At various stages of my work I have enjoyed the collaboration

Finally, I would like to thank most warmly all those ~n Erice,

onW~ occ~ion 06~h~ 15w anniv~any 06~h~

The Hhys of Subnuclear Physics 1

Why Is There Charm, Strangeness, Colour

Do Mesons Fill SU(3) Nonets? 159

The Properties of Charmonium and Charm

Recent Results from DASP 357

New Particles or "Why I Believe in Quarks" 395

New Particle Production in Hadronic Interactions 435

Review of Lepton Production in Hadron - Hadron

Narrow Resonances in BB Reactions 533

Parton Distributions and Their Q2 Dependence 581

Total Cross Sections of Neutrinos and

Measurement of Neutral Current Cross-

Charged V+A Currents in Left-Right Symmetric

Quark and Lepton Mixing 691

Quark-Geometrodynamics: A New Approach to

The Uses of Instantons 805

Can ~le Make Sense Out of "Quantum

Should We Believe in Quantum Field Theory? 983

An Exact Relativistic S-Matrix in 1+1 Dimensions:

Dynamical Symmetries in Nuclear Physics 1043

Symmetries of Quarks and Leptons 1059

The Best Why . 1165

Status of the Subnuclear ~fuys 1171

List of Prizes Awarded, Scholarships, etc. 1215

List of Participants 1217

In the last few years an impressive series of spectacular re-

In spite of these spectacular results, subnuclear physics is

This &s testified by the large number of problems which arise

The purpose of this lecture is to review these problems. Let

i) Why is isospin such a good symmetry? If there are four

ii) Why do degenerate isoscalar-isovector doublets occur fre-

iii) Why are the pseudoscalar mesons so different, with no

i) Why are total cross-sections for strange particles smal-

iii) Why are S- and P-waves in non-leptonic strange particle

Strange particles behave differently from non-strange particles

3. DO WE UNDERSTAND THE NEW PARTICLES?

i) There are four states with C = +1, between J and ~'. Do

iv) Charmed mesons, D, DO, DO, D, have been observed,

g2 (DD):g 2 (DD ):g 2 (D D~< ) ~ 1:5:10.0.